German Yearbook of International Law / Jahrbuch für Internationales Recht: Vol. 60 (2017) [1 ed.] 9783428556588, 9783428156580

Citation preview

VOLUME 60 · 2017

DUNC K ER & HUMBLO T · BERLIN

G E R MAN YEAR B O O K O F I NTE R NATI O NAL LAW Volume 60 · 2017

LIST OF PEER REVIEWERS EBRAHIM AFSAH • University of Copenhagen TILMANN ALTWICKER • University of Zurich

PABLO KALMANOVITZ • Centro de Investigación y Docencia Económicas

DANIELE AMOROSO • Università degli Studi di

GERHARD KEMP • Stellenbosch University

Cagliari HELMUT AUST • Freie Universität Berlin SIGRID BOYSEN • Helmut Schmidt Universität der Bundeswehr Hamburg YASSIN BRUNGER • Queen’s University Belfast JUTTA BRUNNÉE • University of Toronto EMILIANO BUIS • University of Buenos Aires JEAN-YVES DE CARA • Université Paris Descartes Sorbonne Paris Cité STUART CASEY-MASLEN • Geneva Academy of International Humanitarian Law and Human Rights RACHEL CHAMBERS • University of Essex PHILIPP DANN • Humboldt University Berlin SHAI DOTHAN • University of Copenhagen JAN EIJSBOUTS • Maastricht University

JAN KLABBERS • University of Helsinki THOMAS KLEINLEIN • Goethe-Universität Frankfurt am Main VLADISLAV LANOVOY • International Court of Justice FRÉDÉRIC MÉGRET • McGill University CLAIRE METHVEN O’BRIEN • Danish Institute for Human Rights ALEXANDER ORAKHELASHVILI • University of Birmingham CHRISTOPHE PAULUSSEN • T.M.C. Asser Institute MEHRDAD PAYANDEH • Bucerius Law School Hamburg ANNE PETERS • Max Planck Institute for Comparative Public Law and International Law Heidelberg NIELS PETERSEN • Westfälische WilhelmsUniversität Münster

ANDREA GATTINI • University of Padova

ISOBEL ROELE • Queen Mary University of London

TOM GRANT • University of Cambridge

SURYAPRATIM ROY • Trinity College Dublin

JOANNA HARRINGTON • University of Alberta KAREN HULME • University of Essex

MARGOT SALOMON • London School of Economics

REECE JONES • University of Hawai‘i at

NAHED SAMOUR • University of Helsinki

Mānoa

TULLIO TREVES • University of Milan

GERMAN YEARBOOK OF INTERNATIONAL LAW JAHRBUCH FÜR INTERNATIONALES RECHT

Volume 60 · 2017

DUNCKER & HUMBLOT / BERLIN

Founders: RUDOLF LAUN / HERMANN VON MANGOLDT Editors: ANDREAS VON ARNAULD / KERSTIN VON DER DECKEN / NELE MATZ-LÜCK Honorary Editor: JOST DELBRÜCK Guest Editors: ISABELLE HASSFURTHER / WIEBKE STAFF / JENS THEILEN Assistant Editors: DALEY BIRKETT / WIEBKE STAFF Editorial Assistants: MAREIKE NÜRNBERG / DAVID SCHENK / CATHARINA UEKERMANN Layout and Production: ANDREA NEISIUS / SYLVIA WEIDENHÖFER

ADVISORY BOARD OF THE WALTHER SCHÜCKING INSTITUTE CHRISTINE CHINKIN London School of Economics JAMES CRAWFORD International Court of Justice, The Hague

ALLAN ROSAS Court of Justice of the European Union, Luxemburg

LORI F. DAMROSCH Columbia University

BRUNO SIMMA Iran-United States Claims Tribunal, The Hague

RAINER HOFMANN University of Frankfurt

DANIEL THÜRER University of Zürich

FRED L. MORRISON University of Minnesota

CHRISTIAN TOMUSCHAT Humboldt University of Berlin

EIBE H. RIEDEL Geneva Academy of International Humanitarian Law and Human Rights

RÜDIGER WOLFRUM Max Planck Foundation for International Peace and the Rule of Law, Heidelberg

The views presented in the German Yearbook of International Law are those of the contributors and do not reflect or represent the views of the Walther Schücking Institute or the editors, assistant editors, members of the advisory board, or the peer reviewers. Walther Schücking Institute for International Law, University of Kiel Westring 400, D-24098 Kiel, Germany Internet: www.gyil.org

All rights reserved. No part of this book may be reproduced, translated, or utilised in any form or by any means, electronic or mechanical, without the expressed written consent of the publisher. © 2018 Duncker & Humblot GmbH, Berlin Printed by Druckteam, Berlin Printed in Germany ISSN 0344-3094 ISBN 978-3-428-15658-0 (Print) ISBN 978-3-428-55658-8 (E-Book) ISBN 978-3-428-85658-9 (Print & E-Book)

○

Printed on non-aging resistant (non-acid) paper according to ISO 9706 ∞ Internet: http://www.duncker-humblot.de

TABLE OF CONTENTS OBITUARY INGO WINKELMANN: Antonius “Tono” Eitel † (1933–2017) . . . . . . . . . . . . . . . . . . . . . .

11

FORUM The Relationship between African States and the International Criminal Court GERHARD WERLE AND MORITZ VORMBAUM: African States, the African Union, and the International Criminal Court : A Continuing Story. . . . . . . . . . . . . . . . . . . . . . . . .

17

DIRE TLADI: Of Heroes and Villains, Angels and Demons: The ICC-AU Tension Revisited . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

43

FOCUS International Law and the Dehumanisation of Activities HELMUT PHILIPP AUST: “The System Only Dreams in Total Darkness”: The Future of Human Rights Law in the Light of Algorithmic Authority . . . . . . . . . . . . . . . . . . . . . .

71

THOMAS BURRI: International Law and Artificial Intelligence. . . . . . . . . . . . . . . . . . . . . .

91

ALDO CHIRCOP: Testing International Legal Regimes: The Advent of Automated Commercial Vessels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109 STEPHAN HOBE AND BENJAMYN I. SCOTT: International Civil Aviation and the Dehumanisation of Activities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143 STEFAN A. KAISER: Legal Challenges of Automated and Autonomous Systems. . . . . . .

173

6

TABLE OF CONTENTS

NICHOLAS TSAGOURIAS AND RUSSELL BUCHAN: Automatic Cyber Defence and the Laws of War. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203 ANTJE VON UNGERN-STERNBERG: Artifical Agents and General Principles of Law. . .

239

Walther Schücking Lecture PHILIP ALLOTT: Beyond War and Diplomacy: A Giant Step for Mankind. . . . . . . . . . .

269

Special Section Towards Utopia – Rethinking International Law JENS T. THEILEN, ISABELLE HASSFURTHER, AND WIEBKE STAFF: Guest Editors’ Introduction: Towards Utopia – Rethinking International Law . . . . . . . . . . . . . . . . . . . . . . 315 JENS T. THEILEN: Of Wonder and Changing the World: Philip Allott’s Legal Utopianism 335 KA LOK YIP: What is Human? Reading Social Idealism against the Reality of Blackman and Azaria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 365 RADHIKA JAGTAP: Resistance through Utopia: Reflections on the Niyamgiri AntiMining Movement and International Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 393 WIEBKE STAFF: Customary International Law: A Vehicle on the Road from Istopia to Eutopia? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 423 ISABELLE HASSFURTHER: Transforming the “International Unsociety”: Towards Eutopia by Means of International Recognition of Peoples’ Representatives . . . . . . . . . . . 451 DOROTHY MAKAZA: Towards Afrotopia: The AU Withdrawal Strategy Document, the ICC, and the Possibility of Pluralistic Utopias . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

481

SEVERIN MEIER: The Influence of Utopian Projects on the Interpretation of International Law and the Healthy Myth of Objectivity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 515 MARNIE LLOYDD: Persistent Tensions? International Legal Perspectives on ‘Other’ Foreign Fighters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 539 MICHELLE STAGGS KELSALL: From a Stark Utopia to Everyday Utopias . . . . . . . . . . . .

575

TABLE OF CONTENTS

7

ROSSANA DEPLANO: Building Pragmatic Utopias: The “Other” Security Council, International Law, and the United Nations Dream . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 607

GENERAL ARTICLES PETER LAWRENCE AND LUKAS KÖHLER: Representation of Future Generations through International Climate Litigation: A Normative Framework . . . . . . . . . . . . . . . . . . . . . 639 ANJA SEIBERT-FOHR: From Complicity to Due Diligence: When Do States Incur Responsibility for Their Involvement in Serious International Wrongdoing? . . . . . . . . 667

GERMAN PRACTICE AVRIL RUSHE: Same-Sex Marriage under the Grundgesetz and the European Convention on Human Rights . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 711 ISABELLE HASSFURTHER: Will There Be “Justice for Syria”? The Assad Regime in German Courts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 731 FELIX WÜRKERT: The German Past between Collectives and Individuals . . . . . . . . . . .

751

TOBIAS THIENEL: Application and Repeal of the Offence of Insulting Foreign Heads of State: The Böhmermann Affair. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 763 ALENA KUNSTREICH: Prohibition or Non-Proliferation? Germany’s Point of View Concerning the Treaty on the Prohibition of Nuclear Weapons and Effective Nuclear Arms Control and Disarmament . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 773 MARKUS GENTZSCH AND MARC BECKER: PSPP: Curtain Up for a New Act in the Drama “German Federal Constitutional Court versus European Court of Justice”. . . . . 785

8

TABLE OF CONTENTS

BOOK REVIEWS ONUMA Yasuaki: International Law in a Transcivilizational World (TOMUSCHAT)

801

Andrzej Jakubowski/Karolina Wierczyńska (eds.): Fragmentation vs the Constitutionalisation of International Law: A Practical Inquiry (KRZAN) . . . . . . . . . . . . . . . . . . . . . . . 804 Rosalyn Higgins/Philippa Webb/Dapo Akande/Sandesh Sivakumaran/James Sloan: Oppenheim’s International Law: United Nations (KRZAN) . . . . . . . . . . . . . . . . . . . . . . . . 806 Nobuo Hayashi/Cecilia M. Bailliet (eds.): The Legitimacy of International Criminal Tribunals (O’KEEFE). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 808 Christine Chinkin/Mary Kaldor: International Law and New Wars (KOLB) . . . . . . . . .

811

Marina Lostal: International Cultural Heritage Law in Armed Conflict: Case-Studies of Syria, Libya, Mali, the Invasion of Iraq, and the Buddhas of Bamiyan (GAVIRA DÍAZ). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 816 Brian D. Lepard (ed.): Reexamining Customary International Law (RUSHE) . . . . . . . .

820

Anne Peters: Beyond Human Rights. The Legal Status of the Individual in International Law (ROEDER) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 823 C. J. Jenner/Tran Truong Thuy (eds.): The South China Sea: A Crucible of Regional Cooperation or Conflict-making Sovereignty Claims? (YE). . . . . . . . . . . . . . . . . . . . . . . . . 825 Daniel Bodansky/Jutta Brunnée/Lavanya Rajamani: International Climate Change Law (SCHLACKE AND HUGGINS). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 828 Andreas Kulick (ed.): Reassertion of Control over the Investment Treaty Regime (HOPPE) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 830

OBITUARY

Antonius “Tono” Eitel † (1933–2017) Diplomat and international law expert “Tono” Eitel passed away in Münster on 25 June 2017 at the age of 84. With “Tono”, as all his friends called him at his request, we are bidding farewell to a man who inhabited two worlds: the diplomatic world and the world of international law. He was a recognised authority in both of these worlds and was able to engage people in an inimitable way. Tono Eitel as a diplomat. Tono Eitel joined the Federal Foreign Office in 1963. In the course of his career, his postings took him from Kingston to Berne, from Beirut to New York. Having started out in Kingston, he spent five years in Beirut honing the art of diplomatic survival against the backdrop of civil war. He maintained a particular affection for Lebanon and its people throughout his life. His leadership of the Permanent Mission of the Federal Republic of Germany to the United Nations (UN) represented the culmination of his active professional life. Under his direction, Germany held a non-permanent seat on the UN Security Council for two years (1995-1996), and it was during his tenure that the historic and varied discussion surrounding an expansion of the UN Security Council experienced one of its early highlights of the past twenty years in the form of the Razali plan.1 During this period, Germany was able to secure each and every one of its numerous candidacies for UN offices. Eitel was able to win over and impress difficult partners and rivals on the UN stage with both his sincere interest in the other and his quick-wittedness. When an opponent of new permanent seats on the Security Council once claimed that they did not want to create new permanent members as these would surely be “eternal”, Eitel promptly and cheerfully retorted that there was no danger of this happening as tempo-

1

United Nations General Assembly, Paper submitted by the Chairman of the Open-Ended Working Group On The Question of Equitable Representation on and Increase in the Membership of the Security Council and Other Matters Related to the Security Council, UN Doc. A/51/47 Annex II (1997).

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

12

rary “permanent representatives” (i.e. the heads of mission of the UN Member States) were called “permanent representatives” and not “eternal representatives”. In Germany, Eitel was held in high regard from early on as a member of the chancellery staff under Egon Bahr tasked with preparing the Moscow Treaty in 19702 and the Treaty concerning the Basis of Relations between the Federal Republic of Germany and the German Democratic Republic3 in 1972. He published his experiences of this work, which was highly political and of great importance to German post-war politics, under a pseudonym (Benno Zündorf, Die Ostverträge (1979)). Today, this can certainly be regarded as the definitive work on the subject. Subsequent important staging posts in his career included the leadership of a task force working on the Third UN Conference on the Law of the Sea in Jamaica, and finally Director-General for Legal Affairs at the Federal Foreign Office (1992–1995). In this last role, he served as an advisor on international law to the Federal Government and chaired the Federal Foreign Office’s exclusive Advisory Council on Public International Law. Tono Eitel also earned great respect as an international lawyer. He was, to the very last, an External Academic Member of the Max Planck Institute for Comparative Public Law and International Law. He held an honorary professorship at the RuhrUniversität Bochum from 1991. The fact that a special publication was dedicated to him on his 70th birthday (Jochen Abraham Frowein et al. (eds.), Verhandeln für den Frieden/Negotiating for Peace: Liber Amicorum Tono Eitel (2003), 857 pages) is highly unusual for a career diplomat. Among other things, almost three dozen of his own publications on international law are listed therein. Even before his leading role in what came to be known as the Maastricht proceedings before the Federal Constitutional Court in 1993, Eitel was extremely well connected in the German legal scene. His profound humanistic education and outstanding intelligence manifested themselves in an ostentatious modesty, always carefully controlled demeanour and appropriate, expert contributions to what was going on around him. Many German candidates for important international offices owe their success to his support.

2

Treaty between the Federal Republic of Germany and the Soviet Union, 12 August 1970, Federal Republic of Germany-USSR, Bundesgesetzblatt (BGBl.) 1972 II, 354. 3 Treaty concerning the Basis of Relations between the Federal Republic of Germany and the German Democratic Republic, 21 December 1972, Federal Republic of Germany – German Democratic Republic BGBl. 1973 II, 421.

OBITUARY

13

Following his active career, Eitel devoted himself to special missions entrusted to him by the Federal Foreign Office, which never forgot how great an asset he and his network were. At the centre of his life was his beloved paternal home in Münster, whose doors were open to numerous visitors from his circle of colleagues and friends throughout the year. Alongside the old and new centre of his life in Münster, Eitel was also, for many years, Rector of the Postgraduate Program in International Affairs. Through this programme, which was organised and financed by the Robert Bosch Stiftung and the German Academic Scholarship Foundation, he advised and mentored twenty young scholarship holders each year in joint activities preparing them for future employment in international organisations. This work was also rewarding for him and earned him further new “fans”. As a diplomat and person, Tono Eitel belonged to that very rare species that seemed to have nothing but friends. On the diplomatic and international scenes, which are not entirely without their fair share of pronounced self-confidence and ego, this is something that is particularly remarkable. He was able to achieve this feat with many great warm-hearted gestures, by reaching out particularly to those who were smaller and weaker – whether people or countries – and by getting involved in and caring about the lives of others. His spontaneous, dry, and unique humour was also legendary, and never came at the expense of others nor was it delivered without considered courtesy. Time and again, Eitel displayed his great skill in defusing tensions arising during discussions with short, pithy remarks, thereby restoring a positive atmosphere for talks. When he bade farewell to New York (in the summer of 1998), observers were amazed to see delegates in the General Assembly chamber rise from their seats and applaud him. Tono Eitel described his personal career in a private print edition, completed one year prior to his death, entitled ‘Schnee von gestern: zu Hause und im Amt geräumt’ (2016, 296 pages), which offered readers even deeper insights into his personality and the factors that determined his life. Eitel’s childhood in pre-war and war-time Germany left its mark. He was deeply rooted in his Westphalian homeland near Münster. His early and strong proclivity for classical education manifested itself in later life when he managed, from time to time, to work metaphors from Greek mythology into speeches or talks – even in responses in the United Nations General Assembly Hall in New York – without ever appearing overly academic in the process. In later life, he im-

14

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

pressed others with the high degree of forbearance with which he approached weaknesses of others, while always being prepared to exercise restraint and self-criticism, something which is also borne out by his notes. With Tono Eitel’s passing, we bid farewell to a great German post-war diplomat who skilfully straddled two worlds. We have also lost an extraordinary friend and a lasting inspiration for all those who, either in a professional or private capacity, were fortunate enough to have known him. INGO WINKELMANN San José

FORUM

THE RELATIONSHIP BETWEEN AFRICAN STATES AND THE INTERNATIONAL CRIMINAL COURT

African States, the African Union, and the International Criminal Court: A Continuing Story*

GERHARD WERLE(( AND MORITZ VORMBAUM(((

ABSTRACT: This article analyses the strained relationship between African States, the African Union, and the International Criminal Court. It starts by scrutinising the allegations of ‘anti-Africa bias’ that the African Union and some African States have voiced towards the International Criminal Court. Then it looks at the threat of a pull-out of certain African States parties from the ICC Statute after Burundi, South Africa, and The Gambia declared in October 2016 that they were planning to withdraw from the Court. Finally, it analyses the Malabo Protocol, an initiative by the African Union which aims to create criminal chambers in the African Court of Justice and Human and Peoples’ Rights, simply put: an ‘African Criminal Court’. Keywords: Africa, International Criminal Court, Universal Jurisdiction, Selectivity, South Africa, Burundi, Al-Bashir, Heads of State Immunity, Malabo Protocol

I. The Beginning – From Honeymoon to Marital Crisis The relationship between Africa and the International Criminal Court (ICC) got off to a promising start.1 At the State conference in Rome in 1998, many African States * The article is based on a piece written by the authors (Afrika und der Internationale Strafgerichtshof, Juristenzeitung 2015, 581–588) and follows up on the ideas developed in it. (( Professor for German and International Criminal Law, Criminal Procedure, and Modern Legal History at Humboldt-Universität zu Berlin, and Director of the South African-German Centre for Transnational Criminal Justice. (((

Professor for Criminal Law, Criminal Procedure, and International Criminal Law at the University of Münster, and Lecturer at the South African-German Centre for Transnational Criminal Justice. 1

Kai Ambos, Expanding the Focus of the African Criminal Court, in: William A. Schabas/Yvonne McDermott/Niamh Hayes (eds.), The Ashgate Research Companion to International Criminal Law (2013), 499, 508; Gerhard Kemp, Taking Stock of International Criminal Justice in Africa: Three In-

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

18

strongly supported the creation of a powerful international criminal court. Senegal was then the first country to ratify the Rome Statute2 (also referred to as the ICC Statute) in 1999. The first situations the ICC dealt with were, indeed, referred to the Court by African States, namely by Uganda, the Democratic Republic of the Congo, and the Central African Republic. With 34 Member States, Africa became the largest regional group of States parties to the ICC Statute. Many Africans have been employed at the ICC, some in leading positions, including the current Prosecutor of the Court, Fatou Bensouda.

A. Background to the Tension: Controversy between African States and the International Criminal Court

However, only a few years after the Court took up its work, this once promising relationship began to deteriorate. Harsh criticism has been voiced against the ICC by the African Union (AU) and by a number of African States – the AU has called on its Member States not to cooperate with the Court, and a mass pull-out of African States parties from the Rome Statute has even been proposed.3 These actions have been accompanied by severe and polemic verbal attacks. Leading African politicians have expressed the view that the ICC is focusing on Africa while, for the sake of political expedience, it deliberately overlooks international crimes perpetrated elsewhere. The former Chairperson of the Commission of the AU, Mr. Jean Ping (Gabon), for example, has characterised the ICC as a “neo-colonial plaything”.4 The former Ethiopian Prime Minister and then Chairperson of the AU, Mr. Hailemariam Desalegn, stated during an AU summit in 2013 that “the intention [of establishing the International ventories Considered, in: Hermanus J. van der Merwe (ed.), International Criminal Justice in Africa (2014), 7, 12 et seq.; Sanji Mmasenono Monageng, Africa and the International Criminal Court: Then and Now, in: Gerhard Werle/Lovell Fernandez/Moritz Vormbaum (eds.), Africa and the International Criminal Court (2014), 13, 14. 2 Rome Statute of the International Criminal Court, 17 July 1998, UNTS 2187, 90 (Rome or ICC Statute). 3 A selection of relevant African Union (AU) Decisions can be found in: Werle/Fernandez/Vormbaum (eds.) (note 1), 255, 257 et seq., 262 et seq., 265 et seq., 268 et seq., 272 et seq., 274 et seq., 277 et seq. 4

See Max du Plessis/Tiyanjana Maluwa/Annie O’Reilly, Programme Paper: Africa and the International Criminal Court, Chatham House, International Law 2013/01, 1 July 2013, 11, available via: https://www.chathamhouse.org/publications/papers/view/193415 (accessed on 27 November 2017).

AFRICAN STATES, THE AU, AND THE ICC

19

Criminal Court], was to avoid any kind of impunity, but now the process has degenerated to some kind of race hunting”.5 In other words, the Court has been accused of racially biased selectivity and of having become a neo-colonial tool used to target Africa and its leaders.

B. Validity of the Criticism

There are, in fact, a number of grounds for this harsh criticism.6 First of all, as a matter of fact, ten out of eleven situations currently under investigation by the ICC7 are located in Africa. In addition, the ICC has accused two African heads of State, namely the Sudanese President Omar Al-Bashir and the Kenyan President Uhuru Kenyatta, of perpetrating crimes under international law.8 The powerful role the United Nations (UN) Security Council plays in the context of the ICC Statute has also been subjected to criticism by the AU: It has accused the Security Council of, among other things, political selectivity in referring cases to the ICC.9 In addition, a number of African leaders have criticised non-African States for their “targeting” of Africans through their domestic legal systems, alongside the ICC prosecutions, alleging that

5 See Anon., African Union accuses ICC of ‘hunting’ Africans, BBC News, 27 May 2013, available at: http://www.bbc.com/news/world-africa-22681894 (accessed on 7 August 2017). The Ugandan President Museveni repeated the claim almost verbatim on 13 December 2014. 6 For a summary and analysis of the criticism see Tanja Altunjan/Aziz Epik, The International Criminal Court in Crisis?, International Law Observer, 17 January 2017, available at: http://www.inter nationallawobserver.eu/2017/01/17/tanja-altunjan-and-aziz-epik-the-international-criminal-court-incrisis (accessed on 14 August 2017). 7 The situations relate to Uganda, the Democratic Republic of the Congo, the Central African Republic (two situations), Darfur/Sudan, Kenya, Libya, Ivory Coast, Mali, Georgia, and Burundi. 8

The Office of the Prosecutor withdrew the charges against Kenyatta on 5 December 2014 due to lack of evidence; the case, therefore, is considered closed until the Prosecutor submits new evidence. See International Criminal Court (ICC), The Office of the Prosecutor, Notice of withdrawal of the charges against Uhuru Muigai Kenyatta, 5 December 2014, available via: https://www.icc-cpi.int/pages/record. aspx?uri=1879204 (accessed on 7 August 2017). An arrest warrant against Libya’s then head of government, Muammar Gaddafi, was withdrawn on 22 November 2011 after he had been killed. 9 See, for example, the AU Decision of February 2009, in: Werle/Fernandez/Vormbaum (eds.) (note 1), 242 et seq. The criticism is also voiced in an internal document of the AU, see AU, Draft 2 – Withdrawal Strategy Document, 12 January 2017, available at: https://www.hrw.org/sites/default/ files/supporting_resources/icc_withdrawal_strategy_jan._2017.pdf (accessed on 7 August 2017).

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

20

they are “abusing” the principle of universality. This, the argument goes, is tantamount to the practice of neo-colonial judicial imperialism.10 On the one hand, these allegations could be refuted on the ground that they are politically motivated. On the other hand, there could be some truth behind them. A closer look at these points of criticism is thus warranted.

1. Racial Bias Against Africans If one looks at the history of international criminal law, the allegation of selectivity is omnipresent. Already the Nuremberg Trial of Major War Criminals (1945 to 1946) was criticised as being an exercise in “victor’s justice”. Critics have argued that the victors were the judges in this trial, and that those who lost the war were the accused sitting in the dock.11 The Soviet invasion of Poland in 1939, a classic case of a war of aggression, was one example of many that led to accusations of selectivity. While the Germans had invaded Poland from the West, the Red Army attacked it from the East.12 However, no prosecution ever took place against the Soviets in this regard. Similar criticism was voiced against the Tokyo Trial before the International Military Tribunal for the Far East (1946 to 1948) where selectivity could be seen on various levels. The accused were Japanese war criminals, but their head of State, the Japanese emperor (Tennō), was absent. This decision was obviously made to facilitate future cooperation with Japan. “We should not touch the emperor; we’ll take other

10 A number of AU Decisions are concerned with the “abuse of the universality principle”, see in: Werle/Fernandez/Vormbaum (eds.) (note 1), 240, 245, 251 et seq., 253, 259 et seq., 270. See also Evelyne Owiye Asaala, Prosecuting Crimes Related to the 2007 Post-Election Violence in Kenyan Courts: Issues and Challenges, in: van der Merwe (ed.) (note 1), 33, 43 et seq.; Florian Jessberger, ‘On Behalf of Africa’: Towards Regionalization of Universal Jurisdiction?, in: Werle/Fernandez/Vormbaum (eds.) (note 1), 155; Chacha Bhoke Murungu, Towards a Criminal Chamber in the African Court of Justice and Human Rights, Journal of International Criminal Justice (JICJ) 9 (2011), 1067, 1069 et seq. 11

On this criticism see, for example, Mahmoud Cherif Bassiouni, International Criminal Investigations and Prosecutions: From Versailles to Rwanda, in: id. (ed.), International Criminal Law, Vol. 3 (2nd ed. 1999), 31, 45; Gerhard Werle/Florian Jessberger, Principles of International Criminal Law (3rd ed. 2014), MN 25 et seq. 12

See George Ginsburgs, Case Study in Soviet Use of International Law: Eastern Poland in 1939, American Journal of International Law (AJIL) 52 (1958), 69.

AFRICAN STATES, THE AU, AND THE ICC

21

guys instead” seems to have been the motto.13 The Japanese themselves were critical of (among other things) the fact that the Americans were never prosecuted for war crimes for their use of atomic weapons on Hiroshima and Nagasaki.14 The Nuremberg and Tokyo Trials were followed by the Cold War, an age of nonprosecution, despite the perpetration of atrocious crimes – genocide, crimes against humanity, war crimes, and crimes of aggression – in Korea, Cambodia, Vietnam, Zaire (in particular in Katanga province), Nigeria (Biafra), Ethiopia, and numerous Latin American countries, to name only a few examples. The end of the Cold War saw the end of the Security Council’s blockade and the reactivation of international criminal justice. This renaissance was triggered by the human rights atrocities in the former Yugoslavia and the genocide in Rwanda and led to the creation of the two UN ad hoc tribunals: the International Criminal Tribunal for the former Yugoslavia (ICTY) and the International Criminal Tribunal for Rwanda (ICTR). These tribunals were also frequently confronted with allegations of bias.15 Moreover, the creation of the tribunals was itself in a sense selective because they were the only instances where the Security Council could agree to create such courts. At this point in time, we have not seen, for example, a UN tribunal for Chechnya, Crimea, or Syria. So, yes, there has been and continues to be selectivity in international criminal law. Looking at the historical development of international criminal law, one could perhaps say that the alternatives seem to have been either selective prosecutions of crimes under international law or no such prosecutions at all. The question that this article deals with, however, is whether there is selectivity that has been racially biased against Africans, as has been alleged. In other words, does the fact that ten out of eleven situa-

13 See Neil Boister/Robert Cryer, The Tokyo International Military Tribunal (2008), 65 et seq.; Philipp Osten, Der Tokioter Kriegsverbrecherprozess und die japanische Rechtswissenschaft (2003), 105 et seq. 14

See Timothy Brook, The Tokyo Judgment and the Rape of Nanking, Journal of Asian Studies 60 (2001), 673, 694; Osten (note 13), 55, 125. See also Neil Boister, The Tokyo Military Tribunal: A Show Trial?, in: Morten Bergsmo/Cheah Wui Ling/Yi Ping (eds.), Historical Origins of International Criminal Law, Vol. II (2014), 29. 15 See, for example, in International Criminal Tribunal for the former Yugoslavia, Trial Chamber, The Prosecutor v. Slobodan Milosevic, IT-02-54, Decision on Preliminary Motions of 8 November 2001, para. 12; International Criminal Tribunal for Rwanda, Appeals Chamber, The Prosecutor v. Jean-Paul Akayesu, ICTR-96-4, Judgment of 1 June 2001, para. 93; see also Robert Cryer, Prosecuting International Crimes (2005), 209, 220.

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

22

tions the ICC is currently dealing with took or are taking place in Africa indicate political anti-Africa bias on the part of the Court? In a number of situations, this is clearly not the case. Four situations alone were self-referred by States parties from Africa – by Uganda, the Democratic Republic of the Congo, the Central African Republic, and Mali. This means that, in these situations, the African States themselves called upon the ICC to step in. In such cases, one may accuse the referring State party of trying to use the Court to go after the political enemies of the respective government. However, one cannot claim a violation of State sovereignty when the State itself refers a situation to the Court. Two further African situations, namely Sudan/Darfur and Libya, were referred by the UN Security Council. Here, one could accuse the Security Council of political selectivity, but it would be unfair to criticise the ICC for this. Of course, it can be argued that the Court has been used as a tool of the UN where Security Council referrals have taken place. However, in accordance with the ICC Statute, the Prosecutor of the Court has no choice but to open an investigation where a Security Council referral has been made and the general admissibility requirements are fulfilled.16 Four investigations have been opened by the Prosecutor of the ICC proprio motu, namely with regard to Kenya, Ivory Coast, Georgia, and Burundi. The Kenyan situation in particular, in which the later President and his Vice President17 were prosecuted, has sparked criticism against the Court among African leaders. However, one needs to take into account that before the ICC became involved in the Kenya situation, the Waki Commission, which had been tasked to look into the post-election violence of 2007/2008, had proposed, among other things, the establishment of a ‘Special Tribunal for Kenya’ to deal with the alleged crimes.18 The recommendation specifically asserted that a failure to comply with it would result in an ICC interven16

See William A. Schabas, The International Criminal Court (2nd ed. 2016), 375; William A. Schabas/Giulia Pecorella, Art. 13: Exercise of Jurisdiction, in: Otto Triffterer/Kai Ambos (eds.), Commentary on the Rome Statute (3rd ed. 2016), 690, 697. 17

Regarding the charges against President Kenyatta see supra, note 8. The charges against Deputy President William Ruto were vacated following the decision of the Trial Chamber of the ICC on 5 April 2016 that his case be terminated, see ICC, Trial Chamber, The Prosecutor v. William Samoei Ruto and Joshua Arap Sang, ICC-01/09-01/11, Decision on Defence Applications for Judgments of Acquittal of 5 April 2016. This decision was not appealed by the parties. 18 The commission was named after its chairperson, the Kenyan Court of Appeal Judge Philip Waki, see Sosteness Materu, The Post-Election Violence in Kenya (2014), 57 et seq.

AFRICAN STATES, THE AU, AND THE ICC

23

tion. Initially, the Kenyan government signed an agreement to implement the recommendations. However, the Kenyan parliament blocked the implementation process by voting against the establishment of the envisaged domestic mechanism. The political elites, President Kenyatta and Vice President Ruto included, in fact, favoured the ‘Hague option’. Thus, the Kenyan government, by failing to act on the recommendations of the Waki Commission, provoked the intervention of the ICC.19 The Ivory Coast situation is based on an acceptance of the jurisdiction of the ICC. After violence broke out between rebel groups and the military forces of the Laurent Gbagbo government, Ivory Coast, not a party to the ICC Statute at that time, lodged a declaration according to Article 12 (3) ICC Statute on 18 April 2003.20 The declaration was reconfirmed by the new President, Alassane Ouattara, after the postelection violence in 2010.21 On 3 October 2011, the Pre-Trial Chamber granted the Prosecutor’s request to open a proprio motu investigation in the situation in Ivory Coast.22 With regard to the acceptance of the ICC’s jurisdiction by Ivory Coast itself and the reaffirmation of this acceptance under the successor government, one cannot see a case of anti-Africa bias in the Ivory Coast situation. On 25 October 2017, the Pre-Trial Chamber authorised the Prosecutor to open a proprio motu investigation into the situation in Burundi. The Chamber found that there are reasons to believe that State agents, groups implementing State policies, and members of the youth militia Imbonerakure committed crimes against humanity against the civilian population after President Pierre Nkurunziza announced in April 2015 that he was going to run for a third term in office. The ICC has jurisdiction over crimes under the Rome Statute committed in Burundi until 26 October 2017, the date when Burundi’s withdrawal from the ICC took effect.

19

Ibid., 73.

20

See Ivory Coast, Déclaration de reconnaissance de la Compétence de la Cour Pénale Internationale, available at: https://www.icc-cpi.int/NR/rdonlyres/FF9939C2-8E97-4463-934C-BC8F351BA013/ 279779/ICDE1.pdf (accessed on 20 October 2017). 21 See id., Confirmation de la Déclaration de reconnaissance, 14 December 2010, available at: https://www.icc-cpi.int/NR/rdonlyres/498E8FEB-7A72-4005-A209-C14BA374804F/0/ReconCPI. pdf (accessed on 28 November 2017). 22 ICC, Pre-Trial Chamber, Situation in the Republic of Côte d’Ivoire, ICC-02/11-14, Decision Pursuant to Article 15 of the Rome Statute on the Authorisation of an Investigation into the Situation in the Republic of Côte d’Ivoire of 3 October 2011.

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

24

Already on 27 January 2016, the Pre-Trial Chamber granted the Prosecutor’s request to open an investigation into the situation in Georgia with regard to alleged war crimes and crimes against humanity committed in South Ossetia between 1 July and 10 October 2008. It is the first investigation into a situation outside of Africa. In addition, at the time of writing seven out of ten preliminary examinations currently conducted by the Prosecutor relate to non-African States (Afghanistan, Colombia, Iraq/United Kingdom, Palestine, The Philippines, Ukraine, Venezuela).

2. Violation of Heads of State Immunity The ICC has also been harshly criticised for disregarding the immunity of African heads of State. However, a brief look at the ICC Statute shows that immunity for heads of State is excluded by Article 27 ICC Statute. This provision is crystal clear when it says that “official capacity as a head of state or government […] shall in no case exempt a person from criminal responsibility under this Statute” (Article 27 (1) ICC Statute). Therefore, it is substantially unfounded to claim a violation of heads of State immunity if the head of State of a State party is prosecuted by the Court,23 as was the case for Kenya. Admittedly, the legal situation is more complex when it comes to the exercise of jurisdiction against a head of State of a non-State party. This situation arises where the UN Security Council refers a situation to the ICC, as happened in the situations concerning Darfur/Sudan and Libya. There is an ongoing debate as to whether such a referral lifts the immunity of a head of State of the non-State party.24 This is important as States parties would then be under a duty to arrest the head of State of the non-State party and surrender them to the Court – a further cause of disagreement between African States and the ICC.25 Some argue that the UN Security Council must instead explicitly waive the immunity in light of the customary international

23

See, for example, AU Assembly Decision, 26/27 May 2013, AU Doc. Assembly/AU/Dec.482 (XXI), in: Werle/Fernandez/Vormbaum (note 1), 272 et seq. 24 On this debate see, for example, Schabas (note 16), 600 et seq. See also Max Du Plessis/Dire Tladi, The ICC’s Immunity Debate: The Need for Finality, EJIL: Talk!, 11 August 2017, available at: https:// www.ejiltalk.org/the-iccs-immunity-debate-the-need-for-finality/ (accessed on 14 August 2017). 25

See infra, II. B.

AFRICAN STATES, THE AU, AND THE ICC

25

law status of heads of State immunity.26 While this complex question cannot be analysed exhaustively in this article, it is, however, fair to argue that by referring a situation to the Court, the Security Council resolution, indeed, implicitly accepted that the prosecutions would take place according to the ICC Statute’s framework as a whole.27 This implies that a State like Sudan, although not a State party, has obligations under the ICC Statute by virtue of the UN Security Council’s resolution, and Sudan, therefore, is subject to the Court’s jurisdiction and to Article 27 ICC Statute. This mechanism, though far-reaching, makes sense, given that the Security Council has the power to do away with the immunity of heads of State under Chapter VII Charter of the United Nations,28 which has happened in the past, for example, in the case of the ICTY.

3. Reform of the Regulation on the Presence of Accused Persons Before the International Criminal Court One also needs to stress that the States parties to the ICC Statute have addressed some of the concerns of African States. In the context of the Kenya situation, the AU argued that the absence of the head of State and his deputy from their country in order to stand trial in The Hague would have serious consequences for the country’s reconciliation process.29 However, Article 63 (1) ICC Statute clearly requires presence at the Court in The Hague: “The accused shall be present during the trial.” Despite 26 See Dire Tladi, The ICC Decisions on Chad and Malawi, JICJ 13 (2015), 199, 211 et seq. See also the argument by South Africa in ICC, Pre-Trial Chamber, Prosecutor v. Omar Ahmad Al-Bashir, ICC02/05-01/09, Decision under Article 87(7) of the Rome Statute on the Non-compliance by South Africa With the Request by the Court for the Arrest and Surrender of Omar Al-Bashir of 6 July 2017, paras. 34 et seq. 27

See Nerina Boschiero, The ICC Judicial Finding on Non-cooperation Against the DRC and No Immunity for Al-Bashir Based on UNSC Resolution 1593, JICJ 13 (2015), 625 et seq.; Ekaterina Trendafilova, Africa and the International Criminal Court: A Judge’s Perspective, in: Werle/Fernandez/ Vormbaum (note 1), 19, 29 et seq.; Gerhard Werle/Florian Jessberger, Völkerstrafrecht (4th ed. 2016), MN 769. See also ICC, Pre-Trial Chamber, Prosecutor v. Omar Ahmad Al Bashir, ICC-02/05-01/09195, Decision on the Cooperation of the Democratic Republic of the Congo Regarding Omar Al Bashir’s Arrest and Surrender to the Court of 9 April 2014, para. 29. See also ICC, Non-compliance Decision (note 26), para. 85. 28 29

Charter of the United Nations, 26 June 1945, UNCIO 15, 335.

See, for example, AU Assembly Decision, 26/27 May 2013, AU Doc. Assembly/AU/Dec.482 (XXI), in: Werle/Fernandez/Vormbaum (note 1), 272 et seq.

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

26

this clear wording, the States parties agreed in November 2013 to amend the Rules of Procedure and Evidence in a way that the requirement of presence at the trial may also be satisfied by ‘presence’ via video link.30 According to Article 134bis (1) Rules of Procedure and Evidence,31 an accused person “may submit a written request […] to be allowed to be present through the use of video technology […]”. According to Article 134quarter (1) Rules of Procedure and Evidence, such a request is possible if the accused “is mandated to fulfil extraordinary public duties at the highest national level”. It is, therefore, fair to say that the regulation in Article 63 ICC Statute has been substantially extended as a result of the concerns of African States – one could even say overstretched.

4. ‘Abuse’ of Universal Jurisdiction The ‘abuse of universal jurisdiction’ is another accusation by African States that was raised in the context of the criticism against the ICC (although not directly connected with the prosecutions by the Court itself). Universal jurisdiction can, indeed, be exercised across the world with regard to war crimes, genocide, and crimes against humanity, but recent empirical research carried out by Maximo Langer32 has shown no evidence of a racial or geographical bias in the selection of universal jurisdiction cases. In fact, the majority of the accused have been Europeans, namely Nazi perpetrators and Yugoslavs. There is, according to the study, some evidence that universal jurisdiction prosecutions disproportionally affect defendants that are ‘lowcost’ in terms of the political and enforcement resources which their prosecutions require. This group may include Africans, and it has included Rwandans in particular,

30

Assembly of States Parties to the Rome Statute, Resolution ICC-ASP/12/Res.7 of 27 November

2013. 31 Rules of Procedure and Evidence, 3–10 September 2002, Official Records of the Assembly of States Parties to the Rome Statute of the International Criminal Court, First Session (ICC-ASP/1/3 and Corr.1), part II.A. 32 Maximo Langer, The Diplomacy of Universal Jurisdiction, the Role of Political Branches in the Transnational Prosecution of International Crimes, AJIL 105 (2011), 1, 8. For this article, Langer evaluated universal jurisdiction complaints or cases against 1,051 persons (out of which only 32 persons were brought to trial). See also Jessberger (note 10), 165.

AFRICAN STATES, THE AU, AND THE ICC

27

although this is hardly surprising given the worldwide consensus to prosecute and punish those responsible for the Rwandan Genocide.33

C. Interim Conclusion

At this stage, one can say that while there may indeed be some selectivity in the enforcement of international criminal law, the argument that the ICC is ‘targeting’ African States and their leaders is not convincing. The criticism may be understandable, at least to a certain degree, in the context of Africa’s history, its colonisation, and its experience with racism and economic exploitation.34 From a legal point of view, however, the criticism is unfounded. It has also to be stressed that the critical views of the AU and some African leaders are not shared by all African representatives and, in particular, not by the African population, the majority of whom, according to survey findings, support the Court’s actions.35

II. Current Status – Should I Stay or Should I Go? Once the relationship between Africa and the ICC began to deteriorate, some African States proposed a mass pull-out of African States parties from the ICC Statute.36 However, no concrete steps were taken at this time. This changed in October 2016, when Burundi and South Africa submitted their instruments of withdrawal from the ICC Statute to the UN Secretary-General. Shortly thereafter, The Gambia’s then President Jammeh announced the withdrawal of his country – and his Information Minister declared on television that “ICC” has become an acronym for “International Cau33 According to Langer (note 32), 8, 8.28% of the accused persons were from Rwanda, 11.51% from Argentina, 5.23% from the US, 4.19% from China and 4.19% from Israel. 34

On this topic see Res Schuerch, The International Criminal Court at the Mercy of Powerful States, An Assessment of the Neo-Colonialism Claim Made by African Stakeholders (2017). 35 Ambos (note 1), 509; see also Sosteness Materu, A Strained Relationship: Reflections on the African Union’s Stand Towards the International Criminal Court from the Kenyan Experience, in: Werle/ Fernandez/Vormbaum (note 1), 211, 222 et seq., according to whom in surveys conducted in Kenya up to 78% of the interviewees were in favour of the intervention of the ICC in Kenya. 36

See, for example, Anon., Will Africa pull out of the ICC?, BBC News, 11 October 2013, available at: http://www.bbc.com/news/world-africa-24452288 (accessed on 7 August 2017).

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

28

casian Court”.37 Other African States, such as Uganda, welcomed the decisions to withdraw and announced that the mass pull-out of African States would be put back on the AU agenda.38

A. South Africa’s Withdrawal Declaration

In particular, the notice of withdrawal by South Africa – a political heavyweight in Africa and originally one of the strongest supporters of the ICC in Africa – fuelled the ‘ICC vs. Africa debate’. The background to the notice of withdrawal was a visit of Sudanese President Al-Bashir to South Africa in order to attend the AU summit in Johannesburg between 13 and 15 June 2015. On this occasion, the South African government ignored both a request from the ICC to arrest and surrender Al-Bashir, based on the ICC arrest warrants that the Court filed in 2009 against him, and a temporary court order issued by the North Gauteng High Court that stipulated that Al-Bashir not be allowed to leave the country. Instead, the Sudanese President was escorted to a military airport from where he travelled back to his own country.39 South Africa submitted its notification of withdrawal to the UN Secretary-General on 19 October 2016. In this notification, South Africa explained its decision to leave the Court by stating that its “obligations with respect to the peaceful resolution of conflicts at times are incompatible with the interpretation given by the International Criminal Court of obligations contained in the Rome Statute […]”40. This means that, in the view of South Africa, the obligation to arrest Al-Bashir and surrender him to the

37

See Anon., Gambia announces withdrawal from International Criminal Court, Reuters, 26 October 2016, available at: http://www.reuters.com/article/us-gambia-icc-idUSKCN12P335 (accessed on 7 August 2017). 38

See Anon., ICC debate: Africa vs ‘Infamous Caucasian Court’?, Al Jazeera, 28 October 2016, available at: http://www.aljazeera.com/news/2016/10/icc-debate-africa-infamous-caucasian-court-161 028142708060.html (accessed on 7 August 2017). 39

For a summary of the events and the following debates see Tanja Altunjan, Südafrika und der Internationale Strafgerichtshof: Zuma gegen die Gerichte, Junge Wissenschaft im Öffentlichen Recht, 10 March 2017, available at: https://www.juwiss.de/tag/tanja-altunjan/ (accessed on 14 August 2017). 40 Kate Cronin-Furman/Stephanie Schwartz, Is This the End of the International Criminal Court, 21 October 2016, available at: https://www.washingtonpost.com/news/monkey-cage/wp/2016/10/21/ is-this-the-end-of-the-international-criminal-court/ (accessed on 21 August 2018).

AFRICAN STATES, THE AU, AND THE ICC

29

ICC was incompatible with the country’s duty to respect heads of State immunity as reflected in the South African Diplomatic Immunities and Privileges Act 37 of 2001.41 Given that, according to Article 127 ICC Statute, the withdrawal shall take effect one year after the date of receipt of the notification, South Africa was due to lose its status as State party to the Court on 19 October 2017.

B. The Judicial Aftermath

The South African opposition party, the Democratic Alliance, challenged the notice of withdrawal in court. On 22 February 2017, the North Gauteng High Court ruled that the notice was “unconstitutional and invalid”.42 According to the High Court, prior parliamentary approval was required before the notice could be submitted. In reaction to the decision, the South African government revoked its withdrawal from the ICC Statute on 7 March 201743 and has not appealed the High Court’s decision. Yet official representatives have stressed that South Africa still intends to quit the ICC.44 Given that the High Court declared the notice unconstitutional only on formal grounds and did not look into questions of substantive law, submitting a new notice of withdrawal is still possible for the government. In addition, due to its majority in parliament, the government would most probably not face difficulty in gaining a parliamentary majority for the submission of a new notice of withdrawal. In fact, already in December 2017, the South African government introduced an International Crimes Bill in the National Assembly that shall repeal the Implementation of the Rome Statute of the International Criminal Court Act of 2002.45 The bill includes, among others, provisions on crimes against humanity, war crimes, and genocide. In addition, it pro41 Diplomatic Immunities and Privileges Act, No. 37 of 2001, 29 November 2001, Government Gazette of the Republic of South Africa, Vol. 437, No. 22876. 42 High Court of South Africa (Gauteng Division, Pretoria), Democratic Alliance v. Minister of International Relations and Cooperation and Others, Judgment of 22 February 2017, Case No. 83145/2016. 43

UN, South Africa: Withdrawal of Notification of Withdrawal, C.N.121.2017.TREATIESXVIII.10 (Depositary Notification). 44

In December 2017, at the 16th session of the Assembly of States Parties to the ICC, South Africa’s Minister of Justice, Tshililo Michale Masutha, reaffirmed the intention to withdraw from the ICC, available at: https://asp.icc-cpi.int/iccdocs/asp_docs/ASP16/ASP-16-ZA.pdf (accessed on 30 December 2017). 45

Republic of South Africa, International Crimes Bill, available at: http://www.justice.gov.za/legis lation/bills/2017-b37-ICBill.pdf (accessed on 30 December 2017).

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

30

vides for immunity for persons who are immune from criminal jurisdiction of the courts of South Africa under the Diplomatic Immunities and Privileges Act of 2001 (including heads of State or government). The incident during the AU Summit in July 2015 had a legal aftermath at both the national and international level. The North Gauteng High Court46 and the Supreme Court of Appeal of South Africa47 ruled that the South African government had acted unconstitutionally by ignoring the court order and letting Al-Bashir depart to Sudan. The ICC for its part looked into a possible case of non-compliance by South Africa with regard to its duties in terms of the ICC Statute. According to Article 87 (7) ICC Statute, the Court may make a finding of non-compliance and refer the matter to the Assembly of States Parties or the UN Security Council. In its decision of 6 July 2017,48 the ICC ruled that South Africa did, indeed, fail to comply with its obligations under the ICC Statute by not arresting and surrendering Al-Bashir while he was in South Africa. The majority of the judges stressed that by way of the Security Council referral of the situation concerning Darfur/Sudan,49 the Statute as a whole is applicable, including Article 27 (2) ICC Statute, which declares the heads of State immunity irrelevant.50 Therefore, according to the Court, South Africa was obliged to arrest Al-Bashir and surrender him to the ICC. However, the Court did not see sufficient grounds to refer the matter to the Assembly of States Parties or the UN Security Council.

46

The High Court of South Africa (Gauteng Division, Pretoria), The Southern Africa Litigation Centre v. The Minister of Justice and Constitutional Development and Others, Judgment of 24 June 2015, Case No. 27740/2015. 47

The Supreme Court of Appeal of South Africa, The Minister of Justice and Constitutional Development v. The Southern African Litigation Centre, Judgment of 15 March 2016, Case No. 867/15. 48

ICC, Non-compliance Decision (note 26).

49

SC Res. 1593 of 31 March 2005.

50

See also supra, I. B. 2. In his minority opinion, Judge Brichambaut, who agreed with the decision of the majority of the judges, argued that the fact that Al-Bashir did not enjoy immunity rather resulted from “a literal and contextual interpretation of article IV of the Genocide Convention, in conjunction with an assessment of the object and purpose of this treaty”, ICC, Non-compliance Decision (note 26), Minority Opinion of Judge Marc Perrin de Brichambaut, para. 100.

AFRICAN STATES, THE AU, AND THE ICC

31

C. Outlook

After the heated debates that followed the declarations to withdraw by Burundi, South Africa, and The Gambia, things seemed to calm down. The Gambia, after a change of government in January 2017, gave up on the intention to withdraw from the Court.51 At the latest AU summits, a mass pull-out from the ICC Statute was not put on the agenda. The dropping of the charges against Kenyan President Kenyatta and his deputy also contributed to the settling down of the controversy to some extent. Yet, Burundi’s withdrawal took effect on 27 October 2017, and it is still very likely that South Africa will withdraw from the ICC in the near future. The ongoing prosecution of Al-Bashir also remains a cause of disagreement and has the potential to further affect the relationship between the ICC and Africa. Therefore, it has been of great importance that both the ICC and African States parties have made efforts to improve their relationship. During the Assembly of States Parties in November 2016, the Bureau of the Assembly held an open meeting to “engage in a constructive dialogue on the relationship between Africa and the International Criminal Court” and “to seek dynamic solutions”.52 The ICC, in its Non-compliance Decision of July 2017, stressed that South Africa had sought consultation with the Court according to Article 97 ICC Statute before Al-Bashir entered its territory, and that it recognised the “complexity” of the matter for South Africa, thus concluding that the matter should not be referred to the Assembly of States Parties or the UN Security Council.53 Within the AU, a document entitled ‘Withdrawal Strategy Document’ was circulated before the January 2017 summit.54 However, the title of the document is somewhat misleading: Although it does inform African States about the possibility of withdrawing from the ICC Statute, it stresses that there would be ongoing obligations for States parties 51 See the interview with The Gambia’s President Adama Barrow, Deutsche Welle/All Africa, 5 December 2016, available at: http://allafrica.com/stories/201612051105.html (accessed on 14 August 2017). 52

See Assembly of States Parties to the Rome Statute, Informal Summary by the President on the ‘Relationship between Africa and the International Criminal Court’, 24 November 2016, ICCASP/15/36. 53

See ICC, Non-compliance Decision (note 26), paras. 127 et seq. Critical in this regard Angela Mudukuti, Non-Compliance But No Referral – The ICC Muddies the Waters, available at: https://justice inconflict.org/2017/07/20/non-compliance-but-no-referral-the-icc-muddies-the-waters/#more-7386 (accessed on 7 August 2017). 54

See supra, note 9.

32

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

after a withdrawal. In addition, it highlights the fact that a number of African States (explicitly mentioned are Sierra Leone, Ivory Coast, Zambia, Nigeria, Malawi, Senegal, and Botswana) pledged continuous support to the Court after the withdrawal declarations made by Burundi, South Africa, and The Gambia. Whether these developments are the beginning of a better relationship between Africa and the ICC, or whether the relationship will further deteriorate, or whether it will remain in limbo is, at the moment, hard to predict.

III. The Next Chapter – An ‘African Criminal Court’? In connection with the controversy surrounding Africa and the ICC, an Africaspecific project begins to assume a major role not only in the inter-African debate. Ever since the relationship with the ICC began to deteriorate, the AU has adopted repeated resolutions that aim to establish an international, or maybe more precisely, an inter-African criminal jurisdiction for the continent.55 Whereas these resolutions initially attracted scant public attention, and were perhaps even belittled, the situation has now changed: In June 2014, the AU, at its summit in Malabo in Equatorial Guinea, adopted a protocol that creates judicial chambers which shall be empowered to prosecute “international crimes”. The Annex to the Protocol on Amendments to the Protocol on the Statute of the African Court of Justice and Human Rights (Malabo Protocol) contains the Statute which these chambers shall apply.56 Even though the ratification of the Malabo Protocol and its Annex might be a protracted process,57 and some important questions still appear to be unresolved, espe55 Already at its session in February 2009, the Assembly of the African Union requested “the Commission […] to examine the implications of the [African Court of Human and Peoples’ Rights] being empowered to try international crimes such as genocide, crimes against humanity and war crimes […]”, Assembly of the African Union, AU Doc. Assembly/AU/Dec.213(XII), reprinted in: Werle/Fernandez/Vormbaum (note 1), 241. 56 Protocol on Amendments to the Protocol on the Statute of the African Court of Justice and Human Rights, 27 June 2014, available via: https://au.int/en/treaties/protocol-amendments-protocolstatute-african-court-justice-and-human-rights (accessed on 23 November 2017) (Malabo Protocol). 57 According to Art. 11 (1) Malabo Protocol, the “Protocol and the Statute annexed to it shall enter into force thirty (30) days after the deposit of instruments of ratification by fifteen (15) Member States”. So far, there have been ten signatures (by Benin, Chad, Congo, Ghana, Guinea-Bissau, Guinea, Kenya, São Tomé and Príncipe, Sierra Leone, and Uganda) but no ratification of the Protocol, see AU, Status

AFRICAN STATES, THE AU, AND THE ICC

33

cially as regards the financing of the chambers,58 the establishment of an ‘African Criminal Court’ is becoming an increasingly concrete possibility. The Withdrawal Strategy Document of the AU of 2017 encourages African States to ratify the Malabo Protocol and implement it into domestic law, and seeks to identify “regional champions that will work toward driving ratification by member states”.59

A. Content of the Protocol – Overview

Given the comprehensive nature of the protocol, the following sections are restricted to some of its cornerstones and will focus on the crimes the chambers will deal with.60

1. Establishment of the Chambers The ‘African Criminal Court’ consists, in fact, of three criminal chambers. These chambers will be part of the African Court of Justice and Human and Peoples’ Rights. This court itself is not yet operational as its protocol is still in the process of being ratified. Once the protocol has received the sufficient number of ratifications (i.e. fifteen),61 there will be a merger of the African Court of Justice, which is an organ of the AU, and the African Court on Human and Peoples’ Rights, the primary African human rights court.62 Originally, it was planned that this merged Court would have only two sections, with jurisdiction over general issues of the AU as well as over

List, available via: https://au.int/en/treaties/protocol-amendments-protocol-statute-african-court-justiceand-human-rights (accessed on 13 December 2017). 58

According to du Plessis/Maluwa/O’Reilly (note 4), 10, the annual budget of the ICC is fourteen times higher than that of the African Court without the criminal chambers and approximately twice as high as the budget of the AU. See also Murungu (note 10), 1084. 59

See AU, Withdrawal Strategy Document (note 9), 12.

60

For a first comprehensive analysis of the Malabo Protocol and its Annex see Gerhard Werle/Moritz Vormbaum (eds.), The African Criminal Court: A Commentary on the Malabo Protocol (2017). 61 So far, the Protocol has been ratified by Libya, Mali, Burkina Faso, Benin, and the Democratic Republic of Congo. 62

Asaala (note 10), 36 et seq.; Don Deya, Worth the Wait: Pushing for the African Court to Exercise Jurisdiction for International Crimes, Open Society Initiative for Southern Africa (2012), 22, 23.

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

34

human rights and international law. The Malabo Protocol adds a third section with “international criminal jurisdiction” (Article 3 Malabo Protocol).

2. Subject Matter Jurisdiction According to Article 28A Statute of the African Court of Justice and Human and Peoples’ Rights as amended by the Malabo Protocol (the Statute),63 the chambers shall have the power to try persons for the crimes outlined below. First, the Statute includes the four core crimes under international law, which also form part of the ICC Statute.64 As regards the crime of genocide (Article 28B), it is worth mentioning that the Statute adds a new sub-crime, namely “acts of rape or of any other form of sexual violence”. Crimes against humanity under Article 28C now include, in addition to a widespread or systematic attack, a widespread or systematic “enterprise”. As regards war crimes (Article 28D), the Statute follows the ICC Statute in principle, though with some modifications. For example, the crime of conscripting or enlisting child soldiers refers to children under the age of 18 years and not, as is the case in the ICC Statute, to children under the age of 15. Remarkably, the use of nuclear weapons and other weapons of mass destruction is criminalised under Article 28D (g) Statute. The Statute also includes the crime of aggression, adopting largely the provision that was agreed upon at the 2010 Kampala Conference, but also including nonState actors as possible perpetrators of the crime. Although the Malabo Protocol speaks of “international crimes”, the criminal jurisdiction of the chambers goes far beyond the competence of the ICC and includes an impressive list of other crimes of an international or transnational nature. One crime that is conspicuous, for it is unknown in the international context so far, is that of an “unconstitutional change of government” (Article 28E Statute). The inclusion of this crime reflects the large number of coups d’état and attempted coups d’état in Africa and 63

Statute of the African Court of Justice and Human Rights, 1 July 2008, available via: https://au.int/ en/treaties/protocol-statute-african-court-justice-and-human-rights and https://au.int/en/treaties/proto col-amendments-protocol-statute-african-court-justice-and-human-rights (both accessed on 8 December 2017). 64 For a detailed analysis of the provisions on the core crimes in the Statute for the criminal chambers see Kai Ambos, Genocide (Article 28B), Crimes Against Humanity (Article 28C), War Crimes (Article 28D) and the Crime of Aggression (Article 28M), in: Werle/Vormbaum (eds.) (note 60), 31.

AFRICAN STATES, THE AU, AND THE ICC

35

it is modelled on declarations and conventions of the AU, most notably the African Charter on Democracy, Election and Governance of 200765.66 It is important to highlight that Article 28E Statute refers to crimes against a “democratically elected government” and is, therefore, not applicable to dictatorships that came to power via nondemocratic means. Article 28F defines the crime of piracy,67 for which the universality principle has been recognised under customary international law for a long time.68 The crime of terrorism, which has been declared a crime under international law by the Lebanon Tribunal,69 is also included (Article 28G). Here, the definition follows the one contained in the Organization of African Unity Terrorism Convention.70 Article 28H covers the crime of mercenarism, which has also been specifically relevant to Africa.71 Articles 28I and 28Ibis on corruption and money laundering deserve specific attention. The stability of a State is not only at risk from armed conflict and human rights atrocities. It is affected, too, by large-scale economic criminality that has the potential to disrupt the functioning of such States and to deprive the population of natural resources, thus causing famines and other humanitarian catastrophes. It is, therefore, of great interest that the Malabo Protocol adds the crimes of corruption and money

65 African Charter on Democracy, Election and Governance, 30 January 2007, available via: https:// au.int/en/treaties/african-charter-democracy-elections-and-governance (accessed on 23 November 2017). 66

See Gerhard Kemp/Selemani Kinyunyu, The Crime of Unconstitutional Change of Government (Article 28E), in: Werle/Vormbaum (eds.) (note 60), 57; see also Ademola Abass, Prosecuting International Crimes in Africa: Rationale, Prospects and Challenges, European Journal of International Law 24 (2013), 933, 939 et seq. 67 The definition is modelled on Art. 101 United Nations Convention on the Law of the Sea, 10 December 1982, UNTS 1833, 3. See Florian Jessberger, Piracy (Article 28F), Terrorism (Article 28G) and Mercenarism (Article 28H), in: Werle/Vormbaum (eds.) (note 60), 71. 68

See Werle/Jessberger (note 11), MN 133.

69

Special Tribunal for Lebanon, Interlocutory Decision on the Applicable Law, AC, 16 February 2011, paras. 85, 111. See also Matthew Gillett/Matthias Schuster, Fast-track Justice: The Special Tribunal for Lebanon Defines Terrorism, JICJ 9 (2011), 989; Ben Saul, Legislating from a Radical Hague: The United Nations Special Tribunal for Lebanon Invents an International Crime of Transnational Terrorism, Leiden Journal of International Law 24 (2011), 677. 70

OAU Convention on the Prevention and Combating of Terrorism, 14 July 1999, UNTS 2219,

179. 71 The definition of mercenarism follows Article 1 Organization of African Unity Convention for the Elimination of Mercenarism in Africa, 3 July 1977, UNTS 1490, 95. See Jessberger (note 67), 79, 81.

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

36

laundering to the Statute of the African Court.72 However, it rightly includes a limiting element by requiring that the acts of corruption fall under the jurisdiction of the Court only where “they are of a serious nature affecting the stability of a state, region or the union”.73 The Statute also includes trafficking in persons (Article 28J), following the Protocol to the Palermo Convention,74 and trafficking in drugs (Article 28K), following the UN Convention against Illicit Traffic in Narcotic Drugs and Psychotropic Substances75 and the conventions that preceded them.76 While the States parties to the ICC Statute could not agree upon including crimes against the natural environment, the Malabo Protocol does so by way of including trafficking in hazardous wastes and illicit exploitation of natural resources (Articles 28L and 28Lbis).77 Finally, the Protocol makes express provision for the possibility of expanding the list of crimes (Article 28A (2)).

72

See also Arts. 4 and 6 AU Convention on Preventing and Combating Corruption, 11 July 2003, ILM 43 (2004), 5. 73 See Lovell D. Fernandez, Corruption (Article 28I) and Money Laundering (Article 28Ibis), in: Werle/Vormbaum (eds.) (note 60), 89. 74 United Nations Convention against Transnational Organized Crime, 15 November 2000, UNTS 2225, 209. 75 United Nations Convention against Illicit Traffic in Narcotic Drugs and Psychotropic Substances, 20 December 1988, UNTS 1582, 95. 76 Single Convention on Narcotic Drugs, 30 March 1961, UNTS 520, 151, and Convention on Psychotropic Substances, 21 February 1971, UNTS 1019, 175. See Fatuma Mninde-Silungwe, Trafficking in Persons (Article 28J) and Trafficking in Drugs (Article 28K), in: Werle/Vormbaum (eds.) (note 60), 109. 77

The Protocol follows Art. 2 Bamako Convention on the Ban of the Import into Africa and the Control of Transboundary Movement and Management of Hazardous Wastes within Africa, 30 January 1991, UNTS 2101, 177, the African Union’s Protocol Against the Illegal Exploitation of Natural Resources, 30 November 2006, available at: https://ungreatlakes.unmissions.org/sites/default/files/icglr_ protocol_against_the_illegal_exploitation_of_natural_resourcess.pdf (accessed on 7 December 2017), and the African Convention on the Conservation of Nature and Natural Resources (revised version), 11 July 2003, available via: https://au.int/en/treaties/african-convention-conservation-nature-andnatural-resources-revised-version (accessed on 7 December 2017). On the provisions in the Statute as introduced by the Malabo Protocol see Martin Heger, Trafficking in Hazardous Wastes (Article 28L) and Illicit Exploitation of Natural Resources (Article 28Lbis), in: Werle/Vormbaum (eds.) (note 60), 125.

AFRICAN STATES, THE AU, AND THE ICC

37

3. Further Provisions In addition to detailing the crimes over which there shall be jurisdiction, the Statute provides for a far-reaching immunity from prosecution for all incumbent heads of State and government of the AU or those authorised to act in such a capacity, as well as high-ranking, incumbent State officials (Article 46Abis). Article 28N Statute includes a comprehensive provision on modes of responsibility, namely “inciting”, “instigating”, “organizing”, “directing”, “facilitating”, “financing”, “counseling”, “participating as principal, co-principal, agent or accomplice”, “aiding and abetting”, “acting as an accessory before or after the act”, “participating in a collaboration or conspiracy”, and “attempting to commit a crime”. In addition, Article 46C includes corporate criminal liability of legal persons.78 The right to refer a case to the criminal chambers is accorded to States and supraregional bodies as well as to individuals and non-governmental organisations.79 The chambers will be complementary to national courts as well as to the various regional African economic communities.80 In contrast, not a word is mentioned about the ICC.

B. Analysis

The endeavour to create an ‘African Criminal Court’ could be refuted as another political attempt by the AU to weaken the ICC. However, this would ignore the fact that similar initiatives existed in Africa as early as the 1970s, specifically with regard to the apartheid crimes committed in South Africa.81 In order to evaluate the Malabo Protocol adequately, one rather has to draw a distinction between the basic question of the regionalisation of international criminal law on the one hand, and the way the Protocol intends to implement the regionalisation on the other hand. 78

On these provisions see Chantal Meloni, Modes of Responsibility (Article 28N), Individual Responsibility (Article 46B) and Corporate Criminal Liability (Article 46C), in: Werle/Vormbaum (eds.) (note 60), 139. 79

On the provisions on the exercise of jurisdiction see Volker Nerlich, Preconditions to the Exercise of Jurisdiction, in: Werle/Vormbaum (eds.) (note 60), 157. 80

See Harmen van der Wilt, Complementary Jurisdiction (Article 46H), in: Werle/Vormbaum (eds.) (note 60), 187. 81

See Abass (note 66), 936 et seq.

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

38

1. Regionalisation What ‘regionalisation’ of international criminal law means is not clear.82 In a narrow sense,83 one can understand it as adapting international criminal law to the specific needs and circumstances of a certain geographic region. In this sense, we have witnessed the creation of so-called hybrid courts for special situations in different countries – worth mentioning in this context are the Extraordinary African Chambers in the courts of Senegal, which have dealt with the crimes committed by Hissène Habré in Chad.84 The UN ad hoc tribunals, in particular the ICTY for the Balkan region, have also been responsible for specific regions. What has not taken place is regionalisation in the sense of developing a body of international criminal law that is particularly suitable for a specific region of the world. The most typical feature of this kind of regionalisation would be the extension of the catalogue of crimes a regional court is supposed to deal with while retaining jurisdiction over the four core crimes under international law. As we have seen, this is exactly what is provided for in the Malabo Protocol creating the African criminal chambers. Is this kind of regionalisation a retrogressive step? There is no reason it should be. In other legal areas, particularly in the area of human rights, regionalisation has taken place through the adoption of human rights treaties in Europe, America, and Africa without detracting from the core content of the universally accepted human rights; on the contrary, the work of the regional human rights courts has undoubtedly strengthened human rights.85 So why should no similar development be conceivable in international criminal law? The Malabo Protocol does not in any way question the core crimes under international 82

See William W. Burke-White, Regionalization of International Criminal Law Enforcement: A Preliminary Exploration, Texas International Law Journal 38 (2003), 729; William A. Schabas, Regions, Regionalism and International Criminal Law, New Zealand Yearbook of International Law 4 (2007), 3. 83

In a broader sense, the term may also cover the staffing of an international or ‘hybrid’ court with judges from a certain region, Burke-White (note 82), 753. 84 On these Chambers see Mbacké Fall, The Extraordinary African Chambers: The Case of Hissène Habré, in: Werle/Fernandez/Vormbaum (note 1), 117; Murungu (note 10), 1073 et seq. 85

See Mónica Pinto, The Role of the Inter-American Commission and Court of Human Rights in the Protection of Human Rights: Achievements and Contemporary Challenges, Human Rights Brief 20 (2013), 34; Schabas (note 82), 19 et seq.

AFRICAN STATES, THE AU, AND THE ICC

39

law: The elements of the crimes correspond mostly to those contained in the ICC Statute. What the Protocol does do, however, is to include additional crimes which are of special relevance for African States or regions.86 The best example is the crime of unconstitutional change of government. Outside of Africa, this has become a rare phenomenon, even in regions such as South America which for a long time was notorious for its many coups. In contrast, violent overthrows of governments or leaders who remain in their office in violation of the constitution are still a relatively common phenomenon in Africa – countries like Burkina Faso, Burundi, and The Gambia being recent cases in point. In the worst examples, this may destabilise whole region with no possibility of the culprits being prosecuted before the national courts. It was, therefore, only logical that in 2003 the African Charter on Democracy, Election and Governance sought to tackle this problem effectively,87 and declared in Article 25 (5) that “[p]erpetrators of unconstitutional change of government may also be tried before the competent court of the Union”. The Malabo Protocol draws on this initiative. There are further arguments that speak in favour of strengthening regional mechanisms in the field of international criminal law. Regionalisation could in fact mitigate reservations against prosecutions of crimes under international law by international courts or tribunals based on culture. In the case of the Malabo Protocol, one has to take into account that Arusha/Tanzania, where the African Court of Justice and Human and Peoples’ Rights shall have its seat, is, in contrast to The Hague, simply not within the sphere of the former colonial powers. The perception that international criminal justice is a tool of neo-colonial judicial imperialism used by western powers (see supra, I. B. 1.) could, at least to a certain degree, be dispelled with the creation of a regional institution charged with the function of adjudicating crimes under international law. According to Burke-White, such a court is more likely to create a “psychological proximity and a sense of connection between the tribunal and the community”.88 Purely practical concerns also speak in favour of the creation of international criminal chambers or courts at the regional level. While the concept of complementarity 86

Deya (note 62), 24; Murungu (note 10), 1085. For a rather critical perspective see Kemp (note 1),

18. 87

According to the Preamble of the African Charter on Democracy, Election and Governance, coups d’états are “one of the essential causes of insecurity, instability and violent conflict in Africa”. 88

Burke-White (note 82), 736.

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

40

currently distinguishes only between national jurisdiction and the ICC,89 an intermediary court operating at the regional level and acting as a filter mechanism would relieve the ICC of some of its workload, thereby enhancing the effectiveness of international criminal justice.

2. Weaknesses of the Malabo Protocol If the creation of criminal chambers in the African Court of Justice and Human and Peoples’ Rights is then not to be rejected from the start, and if at best it can be seen to enrich the international criminal law landscape, a closer analysis of the Protocol’s contents does raise certain concerns. Clearly, some provisions in the Statute of the new Court are deserving of improvement; the Statute seems to have been drafted in a rush. This becomes apparent, for example, with Article 46Ebis (3), a provision that breaks up in the middle of the sentence and therefore has, so far, “no meaningful content”.90 With regard to the crimes it includes, the Statute frequently incorporates definitions from other international instruments,91 which, however, in many cases were not developed to serve as the basis for prosecutions. The Statute thus has a tendency to ‘over-criminalise’, which is also reflected in Article 28N and its huge number of modes of responsibility. The complementarity requirement upon which the chambers are based (Article 46H) might also lead to problems. According to this provision, the jurisdiction of the Court shall be complementary to that of the national courts, and to the courts of the regional economic communities, such as the East African Economic Community, the Southern African Development Community, etc. The fact of the matter is that the economic communities are based on very different legal foundations and they usually lack the competence to bring criminal charges against an individual. Moreover, many countries belong to several economic communities so that the question arises

89

See Art. 17 ICC Statute.

90

Nerlich (note 79), 171.

91

A juxtaposition of the crimes that the Malabo Protocol introduced and their sources in international instruments can be found in: Werle/Vormbaum (eds.) (note 60), 257 et seq.

AFRICAN STATES, THE AU, AND THE ICC

41

which of these communities is competent if an accused person is a national of such a State or the alleged crimes were committed on its territory.92 The same applies as regards the relationship with the ICC, which is not provided for in the Statute. The most desirable solution would be, as mentioned, the establishment of judicial chambers acting as a filter in support of the ICC. Admittedly, given the basis of the Protocol, this is unlikely for the criminal chambers are, after all, part of an organ of the AU, which, together with one-third of its Member States, is not party to the ICC Statute. In this respect, an integration into the ICC system cannot easily take place.93 Overlapping competencies and unnecessary double expenditure in the prosecution of cases could result, which in the worst-case scenario might lead to increased conflict between Africa and the ICC. Another problematic provision is the immunity clause in Article 46Abis. This provision prescribes immunity for heads of State and heads of government as well as high-ranking officials. Of course, the AU is free to include such a provision in the Statute of its Court, particularly as this does not affect in any way the validity of the immunity clause of the ICC Statute. The problem is rather that it runs counter to the aim of the African Court to fight impunity effectively.94 In other words: Who will be tried if “Head[s] of State or Government” and even “senior state officials” enjoy immunity? And who are “senior state officials” anyway? It will most likely be the lowranking foot soldiers who are envisaged to face trial before the chambers, while the authors, planners, and organisers of gross human rights violations or the beneficiaries of grave economic crimes enjoy immunity under the Malabo Protocol. This provision, therefore, has the potential to ultimately defeat the purpose of the whole project of an ‘African Criminal Court’.

92

See Abass (note 66), 945.

93

See Asaala (note 10), 49; Dire Tladi, Immunities (Article 46Abis), in: Werle/Vormbaum (eds.) (note 60), 203. 94 See also Murungu (note 10), 1082. For an argument against this position, although conceding that the provision is “ambiguous and poorly drafted”, see Tladi (note 93), 211.

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

42

IV. Conclusions We have seen that selectivity in international criminal law is a phenomenon that cannot be denied. In this sense, the criticism of the African Union is substantiated. However, we have also seen that the allegation of a specific anti-African or racial bias is unjustified. As the former Vice President of the International Criminal Court, Ms. Sanji Monageng from Botswana, argues, it is, in fact, absurd that some African States attack the ICC for doing exactly what they – that means the States parties – created the Court to do.95 While the further development of the relationship between African States and the ICC remains uncertain, the hope is that the recent initiatives to revive this relationship will be fruitful and that African States will continue to make effective contributions to fight impunity in Africa in cooperation with the ICC.96 As regards the establishment of chambers in the African Court of Justice and Human and Peoples’ Rights, this article has shown that the regionalisation of international criminal law is not to be viewed negatively, for there are good reasons that speak in favour of it. Admittedly, such a regionalisation should be well thought through; a useful arrangement for criminal courts at the regional level would be to integrate them into the existing system of complementarity between national courts and the ICC. This is not the case with the Malabo Protocol. In addition, we have seen that some provisions are rather poorly drafted. However, one should not be too critical in this regard, especially if we keep in mind that the ICC Statute is not perfect either. With the immunity clause, which exceeds the bounds of an acceptable regionalisation of international criminal law, it is a different story. This aspect of the Protocol represents a drastic step backwards to the period before the Nuremberg trials and it ultimately robs the Protocol of its credibility. Despite this point of criticism, the project of an ‘African Criminal Court’ is a fascinating and, in principle, worthwhile endeavour which, if successful, could serve as a role model for other regions.

95 96

Monageng (note 1), 19.

On African efforts to close the impunity gap see Otilia Anna Maunganidze/Anton du Plessis, The ICC and the AU, in: Carsten Stahn (ed.), The Law and Practice of the International Criminal Court (2015), 65, 68 et seq.

Of Heroes and Villains, Angels and Demons: The ICC-AU Tension Revisited DIRE TLADI(

ABSTRACT: The article re-visits the ICC-AU tension from the lens of the hero-villain trend. It considers the main themes in the ICC-AU debate, namely whether the International Criminal Court (ICC) is targeting Africans and the question of immunity, in the light of the arguments made by the respective interlocutors of the ICC and the African Union (AU). It adopts, as a starting point, the position that the idea that one side is wrong and the other side is right (referred to as the hero-villain approach) is overly simplistic and proceeds to show how complex the issues relating to the ICC-AU tension are. In sum, the article questions both the ICC’s claim to neutrality and the AU’s claim of champion of decolonisation. The article concludes with a postscript to account for the Prosecutor’s recent decision to seek an investigation into the situation in Afghanistan. KEYWORDS: Immunity, ICC-AU, Self-Referrals, Jurisdiction, Non-Cooperation, Al Bashir

I. Introduction When asked by the editors of the German Yearbook of International Law to make a contribution to the discussion on the International Criminal Court (ICC)-Africa relations in this Forum Section, my immediate reaction was that the African-ICC relations question has been largely overshadowed in the last two years by the ICCSouth Africa relations and was thus passé.1 Yet, in many ways the South Africa-ICC ( Professor of International Law, Department of Public Law and Research Fellow Institute of Comparative and International Law in Africa, University of Pretoria. Member of the UN International Law Commission and its Special Rapporteur on the topic of Peremptory Norms of General International Law (Jus Cogens); Special Adviser to the Minister of International Relations and Cooperation of South Africa. The views expressed herein are not attributable to any of the institutions affiliated with the author. 1 The author served as lead counsel for South Africa during its non-cooperation hearing. See International Criminal Court (ICC), Pre-Trial Chamber II, The Prosecutor v. Omar Hassan Ahmad Al Bashir, ICC-02/05-01/09-302, Decision under Article 87(7) of the Rome Statute on the Non-Compliance by

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

44

relationship is a microcosm of the Africa-ICC tensions. I have thus gladly accepted to make a contribution. In previous contributions on this subject, I have questioned the dominant trend in which the debate on the ICC-African Union (AU) relationship is characterised by an ideological chasm that has villains pitted against protagonists.2 In this characterisation, both sides (and their respective interlocutors) describe themselves as heroes fighting the good fight and the other side as evil villains intent on wanton destruction.3 In the context of the ICC-AU debate, the ICC and its interlocutors present themselves as the defenders of accountability and justice while depicting the AU and its interlocutors as promoting impunity. They claim that, in the pursuit of justice, the ICC does not bend to politics and concerns itself with only law.4 The AU and its

South Africa with the Request by the Court of the Arrest and Surrender of Omar Al Bashir of 6 July 2017. For the author’s personal views on the South Africa cases see Dire Tladi, The Duty on South Africa to Arrest and Surrender Al Bashir under International and South African Law: A Perspective from International Law, Journal of International Criminal Justice (JICJ) 13 (2015), 1027; id., Interpretation and International Law in South African Courts: The Supreme Court of Appeal and the Al Bashir Saga, African Human Rights Law Journal 16 (2016), 310. 2

Id., When Elephants Collide it is the Grass that Suffers: Cooperation and the Security Council in the Context of the AU/ICC Debate, African Journal of Legal Studies 7 (2014), 381, 381. 3

Ibid., 381–382. Kelly-Jo Bluen has made a similar observation: “[…] many liberal advocates of the ICC criticise the AU’s conflation of law and politics (yet suspiciously ignore the similar conflation of others) […]”, Kelly-Jo Bluen, Imperfect ICC Can Play a Critical Role, Mail & Guardian Online, 3 February 2017, available at: https://mg.co.za/article/2017-02-03-00-imperfect-icc-can-play-a-critical-role (accessed on 17 August 2017). 4

The idea of the ICC acting only in accordance with the law is reflected in a number of statements by members of the ICC. In a recent article, Judge Sanji Monageng made the following observation: “[…] the Court cannot cure any defects of the Rome Statute – our job is to interpret and apply the Statute. After all, the International Criminal Court is a court, a judicial body that must respect its legal framework. Anything else would be unthinkable for judges”, Sanji Monageng, Africa and the International Criminal Court: Then and Now, in: Gerhard Werle/Lovell Fernandez/Moritz Vormbaum (eds.), Africa and the International Criminal Court (2014) 13, 18. Furthermore, shortly after his election to the position of Prosecutor, Luis Moreno-Ocampo, the first Prosecutor, stated that the “ICC was totally independent and impartial”, Luis Moreno-Ocampo, Statement made at the ceremony for the solemn undertaking of the Chief Prosecutor of the ICC, 16 June 2003. The current Prosecutor, shortly after election, emphasised that the “Prosecutor does not take into account any political considerations”, Fatou Bensouda, Introductory Remarks, The International Conference: 10 Years in Review of the ICC – Justice for All?, 15 February 2012. The first President of the ICC, Philippe Kirsch, suggested that the ICC was a response to the “bargaining away of justice” for “short-term political gain”, Judge Philippe Kirsch, Address to the United Nations General Assembly, 30 October 2008. See for discussion, Sarah Nouwen/Wouter G. Werner, Doing Justice to the Political: The International Criminal Court in Uganda

THE ICC-AU TENSION REVISITED

45

interlocutors, on the other hand, see themselves as the defenders of sovereign equality of States and the ICC and its interlocutors as promoting neo-imperialism and the dominance of the weak by the powerful. They argue that the ICC’s claim of objective application of law is nothing more than a cloak to cover the politically motivated bias of the ICC. My view is that this hero-villain trend is overly simplistic. A dispassionate review of the accounts, decisions, literature, and facts surrounding the ICC will reveal that the ICC-AU debate is far more complex than the hero-villain trend suggests. It will reveal that there are no white hats and black cloaks; it will reveal that what we have are international actors whose eager pursuit of their respective mandates and objectives leads them to adopt particular positions on both law and politics. There are a variety of angles from which the nuances of the ICC-AU relationship can be considered. One angle, not considered in this article but addressed in Gerhard Werle and Moritz Vormbaum’s article,5 concerns the adoption by the AU of the Malabo Protocol to establish a criminal chamber in the African Court, which was, at least in part, fuelled by the tension between the ICC and the AU.6 In this article, I propose to consider the ICC-AU tension from two angles, namely selectivity in investigations and prosecutions and the rules relating to immunity under international law. I have selected these two broad themes because it has been largely through these themes that the ICC-AU tension, and consequently the herovillain trend, has played itself out. But other equally relevant angles from which the tension could be studied are cooperation and complementarity. I begin, in the next section, by considering the first theme – selectivity in prosecutions, or whether the ICC is ‘targeting’ Africa. I consider, in particular, the arguments and counter-arguments made by both sides with the aim of showing that neither side comes to the

and Sudan, European Journal of International Law 21 (2011), 941; David Kaye, Who is Afraid of the International Criminal Court: Finding the Prosecutor Who Can Set it Straight, Foreign Affairs 90 (2011), 11. 5

See Gerhard Werle/Moritz Vormbaum, African States, the African Union, and the International Criminal Court: A Continuing Story, German Yearbook of International Law 60 (2017), 17. 6

Protocol on the Amendments to the Protocol on the Statute of the Africa Court of Justice and Human Rights, 27 June 2014, available at: https://au.int/sites/default/files/treaties/7804-treaty0045_-_protocol_on_amendments_to_the_protocol_on_the_statute_of_the_african_court_of_ justice_and_human_rights_e.pdf (accessed on 7 December 2017).

46

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

table with clean hands.7 I then consider the arguments concerning immunity. Various arguments are made by both sides of the divide. Yet, by and large, both sides ignore important aspects of international law and gloss over the details of the legal rules in order to arrive at the conclusion that justifies their image as white hat-wearing heroes and the other as ‘black-cloaked’ evil villains. I then offer some concluding remarks. A postscript to address the recent decision of the Office of the Prosecutor to request authorisation for opening an investigation into Afghanistan subsequent to the writing of the article is also provided. Since this article is, in some sense, about the objectivity of the players in the ICCAU debate, a word or two about my own objectivity is warranted. I have, on account of my association with the South African position, been presented (mainly in private) of adopting positions to protect South Africa. Yet my position has been fairly consistent, even before South Africa was catapulted into the limelight by the Al Bashir saga. In 2009 I wrote an article admonishing the AU’s ‘neo-colonialism’ rhetoric.8 I make this point to dispel any notion that my views are predetermined by my relationship with the South African government or the AU. My only loyalty, in these discussions, is to the law.

II. The ICC Targets Africa A. General

Perhaps the best starting point for the vexed question of whether the ICC is targeting Africa is a look at the gross numbers: At the time of writing, the ICC is seized with ten situations.9 Of these situations, nine are on the African continent. Indeed, until recently, nine out of nine situations under investigation by the ICC 7 See for a description Tim Murithi, Between Political Justice and Judicial Politics: Charting a Way Forward for the African Union and the International Criminal Court, in: Werle/Fernandez/ Vormbaum (eds.) (note 4), 181. 8 Dire Tladi, The African Union and the International Criminal Court: The Battle for the Soul of International Law, South African Yearbook of International Law 34 (2009), 57. 9 These are Uganda, the Democratic Republic of the Congo, Darfur (Sudan), Central African Republic, Kenya, Libya, Côte d’Ivoire, Mali, Central African Republic (II), and Georgia. Information available at: https://www.icc-cpi.int/pages/situation.aspx (accessed on 24 August 2017).

THE ICC-AU TENSION REVISITED

47

concerned African situations.10 According to the ICC website, the Court is currently seized with 25 cases, all of which are against Africans.11 The proportion of African cases in the ICC is even larger if one includes cases that have been discontinued such as that against Muammar Gaddafi and El Senussi of Libya, and if one considers that some of these cases have multiple accused, so that the number of persons that have been indicted by the ICC stands at 41, all of whom are African. The proportion of situations, cases, and indictments from Africa is quite frankly staggering. These numbers, and the high proportion of Africans reflected in them, are not in dispute. What is in dispute is what these numbers mean. Do these numbers indicate that the Court cares for Africa? Are they rather an indication that the Court is targeting Africans, while permitting impunity elsewhere in the world? Arguments back and forth have been made about the meaning of the empirically incontrovertible high proportion of African situations, cases, and indictees before the ICC. The legal dictum res ipsa loquitor – the facts speak for themselves – would seem to suggest that, unless arguments are advanced to show that there is not a targeting of Africans, then we ought to accept that there is a targeting of Africans. It is useful, therefore, to proceed by identifying the common arguments that have been advanced by the ICC (and its interlocutors) for the lopsided proportion of cases on the African continent. While there are several arguments that have been advanced,12 the main arguments are, first, that most of the situations that the ICC is seized with have been self-referred and, second, that the ICC is constrained by its jurisdictional mandate. 10

The ICC opened an investigation into the situation in Georgia in January 2016.

11

See information available at: https://www.icc-cpi.int/Pages/cases.aspx (accessed on 24 August 2017). While the situation in Georgia has been under ICC investigation since January 2016, to date there has not been a single case opened. 12

Other, more political, arguments include the contention that if we focus more on the victims, then the lopsidedness is indicative of a concern for Africa. See, e.g., Kenneth Roth, Africa Attacks the International Criminal Court, Human Rights Watch, available at: https://www.hrw.org/news/2014/ 01/14/africa-attacks-international-criminal-court (accessed on 24 August 2017), where the following question is posed: “What are we to make of the fact that in its eleven-year history, the International Criminal Court (ICC) has prosecuted only Africans? Should the court be condemned for discrimination – for taking advantage of Africa’s weak global position – as some African leaders contend? Or should it be applauded for giving long-overdue attention to atrocities in Africa –a sign that finally someone is concerned about the countless ignored African victims, as many African activists contend?” This argument concerning a focus on victims does not take us very far. Assuming that crimes within the jurisdiction of the Court are being perpetrated elsewhere and the Court chooses to focus on African victims at the expense of other victims, this would give off a smell of patronisation that I do not think the ICC would want to be associated with – it would suggest that others can take care of themselves but Africa cannot.

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

48

There is an argument that is rarely (if ever) made in writing but that does emerge from time to time. It is sometimes suggested that it is factually incorrect to state that the ICC is only seized with the situations on the African continent because it is also seized with the situation in Afghanistan, Colombia, Iraq, and others.13 The argument, however, obfuscates a very real distinction between situations under investigation and situations under preliminary analysis. The latter are not active investigations. These are merely situations that the Court is keeping an eye on and considering whether to investigate. There are no legal consequences flowing from, and certainly no actual cases concerning these situations. Some of these situations have been under preliminary analysis for around ten years without any actual investigations.14 It is thus disingenuous to suggest that these preliminary analysis can be equated with the Court being seized with a situation.

B. Africans Have Referred the Situations Themselves

The first of the main arguments in response to the claim that the ICC is targeting Africans is that African States have, by and large, referred themselves to the ICC.15 This contention, also made in the article by Werle and Vormbaum, is empirically true. Uganda, Mali, Central African Republic, and Democratic Republic of Congo were all ‘self-referred’ situations.16 The situation in Côte d’Ivoire, which is reflected on the ICC website as one which was initiated by the Prosecutor of her own volition (proprio motu), might also be counted as a ‘self-referral’ because the investigations were begun on account of a declaration by Côte d’Ivoire recognising the Court’s jurisdiction though it was not, at the time, a State party. Yet, it is well known that these ‘self-

13 During a Wayamo Foundation-sponsored South African-German Dialogue held in Pretoria from 27 to 28 March 2017, one participant, who cannot be named as the dialogue took place under the Chatham House Rule, made this assertion. 14

According to the ICC website, for example, the situation in Afghanistan “was made public in 2007” suggesting that it was being considered even before that, cf. https://www.icc-cpi.int/afghanistan (accessed on 24 August 2017). 15 16

See Monageng (note 4), 17.

Although the phrase ‘self-referral’ is en vogue, the Statute itself does not use that phrase. The Rome Statute of the International Criminal Court, 17 July 1998, UNTS 2187, 3 (Rome Statute), only contemplates State referrals, which would include situations in the territory of one State party referred by another State party.

THE ICC-AU TENSION REVISITED

49

referrals’ were, in reality, generated by the Office of the Prosecutor.17 Indeed the Office of the Prosecutor has itself conceded that the Prosecutor encouraged States to refer situations in their own countries.18 While the fact that these self-referrals were pursued by the Office of the Prosecutor does not, in itself, minimise the legal value of the self-referrals and does not suggest any bias, it certainly casts doubt on the image of African States eager for justice running to the ICC to provide that justice. It is also interesting that in all – not some, not many, but all – these self-referred situations, the accused persons that find themselves in the dock are all rebels in conflicts against the governments that made the self-referrals. This has raised the suspicion whether these self-referrals were based on some kind of quid pro quo – the self-referral serves to malign the enemies of the government while raising the legitimacy of the ICC.19 In many ways, the arguments concerning ‘self-referrals’ are neither here nor there. On the one hand, whether the self-referrals are a result of pressure, encouragement, or requests from the ICC, their legal value as State referrals cannot be diminished. On the other hand, even if these ‘self-referrals’ were genuine, it does not change the fact that the Court’s work is disproportionately focused on African States. Discounting self-referred situations (including Côte d’Ivoire), the ICC has, in total, indicted 21 individuals, all of whom are African. The disproportion thus remains extremely high. The real argument in response to the disproportionate number of prosecutions of African persons must, therefore, be the jurisdictional constraints on the ICC. C. Jurisdictional Constraints Prevent the ICC from Looking Elsewhere

Judge Sanji Monageng’s comments, in response to claims that the ICC is targeting Africans, succinctly captures the jurisdictional constraints arguments: “[…] the Rome Statute does not give the Court universal jurisdiction. The Court may only exercise jurisdiction over crimes committed on the territory of a State Party, or by a national of a State Party.”20 17 See, e.g., Sarah Nouwen, Complementarity in the Line of Fire: The Catalysing Effect of the International Criminal Court in Uganda and Sudan (2013), 13; Andreas Müller/Ignaz Stegmiller, SelfReferrals on Trial: From Panacea to Patient, JICJ 8 (2010), 1267. See also Bluen (note 3). 18

Müller/Stegmiller (note 17), 1267.

19

Nouwen (note 17), 116.

20

Monageng (note 4), 15.

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

50

This argument, which is often repeated, suggests that the Court is constrained from opening investigations in other places outside of Africa where Rome Statute of the International Criminal Court (Rome Statute) crimes may have been committed because the Court does not have jurisdiction in those situations. It is certainly true that in some cases where the ICC is criticised for not acting it simply cannot act because it does not have jurisdiction. The calls for the ICC to indict George W. Bush and Tony Blair21 and open investigations into Syria,22 for example, ignore (or are ignorant of) the fact that the ICC would not have jurisdiction in those instances. There is thus certainly some truth to the argument that the ICC is unable to act in some cases outside of Africa. While there are situations outside of Africa in which Rome Statute crimes are being committed but which fall beyond the reach of the ICC, the jurisdictional constraints argument is insufficient to explain the Court’s lack of action outside of Africa. There are key situations that the ICC has chosen not to open investigations in, notwithstanding that it has jurisdiction. The situation in Afghanistan provides the first example of a situation in which serious crimes are known to have been committed23 and in which the ICC has yet to act though it has jurisdiction. Some of the war crimes committed in Afghanistan include possible attacks on civilian installations, such as hospitals.24 The ICC has jurisdiction over the situation in Afghanistan be21 Thomas Siepelmeyer, Can the ICC Prosecute Tony Blair and Others Who Can Still Evade Prosecution Today, Alternative Information & Development Centre, 23 November 2016, available at: http://aidc.org.za/can-icc-prosecute-tony-blair-others-can-still-evade-prosecution-today/ (accessed on 25 August 2017). 22 On possible ICC action in Syria see Mark Kersten, Calls to Prosecute War Crimes in Syria Are Growing: Is International Justice Possible?, Justice in Conflict, 17 October 2016, available at: https:// justiceinconflict.org/2016/10/17/calls-to-prosecute-war-crimes-in-syria-are-growing-is-internationaljustice-possible/ (accessed on 25 August 2017). See also Dire Tladi, ICC and UNSC: Point Scoring and the Cemetery of Good Ideas, ISS Today, 10 October 2014, available at: https://issafrica.org/iss-today/ icc-and-unsc-point-scoring-and-the-cemetery-of-good-intentions (accessed on 13 November 2017). 23 Anon., US Army and CIA may be Guilty of War Crimes in Afghanistan, says ICC, The Guardian, 15 November 2016, available at: https://www.theguardian.com/law/2016/nov/15/us-army-and-ciamay-be-guilty-of-war-crimes-afghanistan-says-icc (accessed on 26 August 2017). 24

See, e.g., Spencer Ackerman/Sune Engel Rasmussen, Kunduz Hospital Attack: MSF’s Questions Remain as US Seeks no Charges, The Guardian, 29 April 2016, available at: https://www.theguardian. com/us-news/2016/apr/29/kunduz-hospital-attack-msf-us-military-charges (accessed on 27 August 2017). See also, covering earlier cases of possible war crimes, Human Rights Watch, “Troops in Contact”: Airstrikes and Civilian Deaths in Afghanistan, US and NATO Bombing and Civilian Deaths, September 2008, 12, available at: https://www.hrw.org/sites/default/files/reports/afghanistan0908 web_0.pdf (accessed on 13 November 2017). It has been reported that between 2007 and 2011 “10.000

THE ICC-AU TENSION REVISITED

51

cause Afghanistan is a party to the Rome Statute. Indeed, the ICC has had the situation under preliminary analysis since at least 2007. Given the clear jurisdictional competence and the undeniable commission of war crimes, it is unclear why, in more than ten years, the ICC has not opened an investigation, let alone cases, in the situation in Afghanistan. In 2016, concluding its report on the situation in Afghanistan, the Office of the Prosecutor declared that it “will make a final decision on whether to request a Pre-Trial Chamber authorisation to commence an investigation into the situation” in Afghanistan.25 The Office of the Prosecutor, however, has yet to decide on an investigation and has, in fact, announced that it would be delaying any decision on an investigation.26 Another situation in which the ICC could have, but has not, acted is the situation in Iraq. Iraq, of course is not a State party to the Rome Statute. However, the ICC can exercise jurisdiction over crimes committed by nationals of States parties for crimes committed on the territories of non-States parties.27 It is well documented that United Kingdom (UK) soldiers – the UK is a State Party – have committed war crimes in Iraq.28 The Chilcot Report, for example, noted that accounts of the death toll in Iraq since the invasion of the Coalition forces are unclear, but what “is not in doubt is that, in both the military operation to overthrow the Iraqi regime and the subsequent violence, many tens of thousands of Iraqi citizens, most of them civilians,

civilians have been killed in the conflict in Afghanistan”, Office of the Prosecutor, 2013 Report on Preliminary Examination Activities, para. 24. 25

Office of the Prosecutor, 2016 Report on Preliminary Examination Activities, para. 230.

26

Stephanie van den Berg, ICC: New Information Delays Decision on Afghan War Crimes Investigation, Reuters, 5 July 2017, available at: https://uk.reuters.com/article/uk-warcrimes-afghanistan-icct/ icc-new-information-delays-decision-on-afghan-war-crimes-investigation-idUKKBN19Q1ZW (accessed on 27 August 2017). 27 28

Art. 12 (2)(b) Rome Statute.

Kevin Rawlinson, Nearly 300 British Veterans Face Investigations over Alleged Iraq War Crimes, The Guardian, 9 January 2016, available at: https://www.theguardian.com/uk-news/2016/jan/09/ nearly-300-british-veterans-face-investigation-over-alleged-iraq-war-crimes (accessed on 27 August 2017); Tom Batchelor, British Troops Must Face Investigations for War Crimes in Iraq, Insists Corbyn, Express, 25 September 2016, available at: http://www.express.co.uk/news/uk/714165/Jeremy-CorbynBritish-Troops-Must-Face-Investigation-Over-Alleged-Abuses-Iraq (accessed on 27 August 2017). See also Clive Baldwin, Time to Properly Investigate UK War Crimes in Iraq: The British Military Justice System Has Shown Itself Unfit for Purpose in Handling War Crimes Committed Outside the Country, The Independent, 16 January 2014, available at: https://www.hrw.org/news/2014/01/16/time-properlyinvestigate-uk-war-crimes-iraq (accessed on 27 August 2017).

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

52

lost their lives”.29 It is also clear that the UK does not, at least for the moment, have any intention of investigating and prosecuting these alleged cases of war crimes itself.30 It is also worth pointing out that, as with the situation in Afghanistan, the situation in Iraq has been under preliminary analysis by the Office of the Prosecutor for more than ten years.31 The situation in Palestine is another situation over which the ICC would have jurisdiction to act and which has been under preliminary analysis for nearly a decade. In 2012, the previous Prosecutor concluded, after a preliminary analysis spanning a period of three years, that the ICC could not exercise jurisdiction over the situation, notwithstanding Palestine’s deposit of a unilateral declaration under Article 12 (3) Rome Statute, on the ground that Palestine was not a State.32 On obtaining the collective recognition of its statehood by the United Nations General Assembly through the observer State status,33 Palestine made another unilateral declaration and then acceded to the Rome Statute, becoming a party. The ICC thus has jurisdiction. Again, the Court, after many years of preliminary analysis, has not opened an investigation. Again, the existence of crimes against humanity, including the crime of apartheid, and war crimes seems beyond doubt.34

29 Sir John Chilcot (ed.), The Report of the Iraq Inquiry: Report of a Committee of Privy Counsellors (2016), 170. 30

Damien Gayle, May Vows to Protect UK Troops Who Fought in Iraq from Legal Abuse, The Guardian, 23 September 2016, available at: https://www.theguardian.com/uk-news/2016/sep/23/ theresa-may-british-troops-uk-protect-abuse-legal-system-soldiers-war-crimes-iraq (accessed on 27 August 2017); see also Robert Mendick, Outrage as War Crimes Prosecutors Say Tony Blair will not be Investigated Over Chilcot’s Iraq War Report – But British Soldiers Could be, The Telegraph, 2 July 2016, available at: http://www.telegraph.co.uk/news/2016/07/02/outrage-as-war-crimes-prosecutorssay-tony-blair-will-not-be-inv/ (accessed on 27 August 2017). 31

See Luis Moreno-Ocampo, former Prosecutor, unaddressed letter of 9 February 2006, available at: https://www.icc-cpi.int/NR/rdonlyres/04D143C8-19FB-466C-AB77-4CDB2FDEBEF7/143682/ OTP_letter_to_senders_re_Iraq_9_February_2006.pdf (accessed on 27 August 2017). 32

David Davenport, International Criminal Court Prosecutor Resists Palestinian End-Run, Forbes, 23 April 2012, available at: https://www.forbes.com/sites/daviddavenport/2012/04/23/internationalcriminal-court-prosecutor-resists-palestinian-end-run/#27927808660c (accessed on 27 August 2017). 33 34

See General Assembly Res. 67/19 of 4 December 2012 (Status of Palestine in the United Nations).

See William Schabas, International Criminal Law and the Middle East, in: Tiya Maluwa/Max du Plessis/Dire Tladi (eds.), The Pursuit of a Brave New World: Essays in Honour of John Dugard (2017), 472; see also Office of the Prosecutor, 2015 Report on Preliminary Analysis, paras. 62–71. See further Human Rights Watch, Palestine: ICC Should Open Formal Probe, 5 June 2016, available at: https:// www.hrw.org/news/2016/06/05/palestine-icc-should-open-formal-probe (accessed on 17 August 2017).

THE ICC-AU TENSION REVISITED

53

These situations illustrate that the jurisdictional constraints argument is simply not true. What they do not tell us, though, is why the Office of the Prosecutor has, after all this time, failed to make decisions on each of these situations. The annual reports of the Office of the Prosecutor on these issues do not offer much assistance either. With respect to the situation of UK war crimes in Iraq, for example, the 2016 report states simply that the Office of the Prosecutor “is currently in the process of concluding its comprehensive factual and legal assessment of information available” so as to “establish whether there is a reasonable basis to believe that alleged crimes committed by United Kingdom nationals” in Iraq between 2003 and 2009 “fall within the subject-matter jurisdiction of the Court.”35 This ‘reason’ is similarly reflected in the reports on the situations in Afghanistan and Palestine.36

D. The Role of Power

This begs the question why, given the existence of so much material evidencing crimes in all three situations, it has taken so long for the Office of the Prosecutor to commence actual investigations. The length of time to consider whether there is reason to launch investigations stands in stark contrast to the speed and eagerness to open investigations and issue indictments (and arrest warrants) in situations involving Africans. In the situation in Libya, for example, the jurisdiction was conferred on the ICC by virtue of United Nations Security Council (UNSC) Resolution 1970 on 26 February 2011. Barely five days later (not years, months, or weeks, but days), on 3 March 2011, the Office of the Prosecutor had conducted and completed its preliminary analysis to determine that there is a reasonable basis to believe that crimes under the ICC’s jurisdiction had been committed. Eleven, ten, and eight years respectively in Iraq, Afghanistan, and Palestine. Is it not reasonable to conclude – as I have – that the Prosecutor genuinely wants to open investigations in these situations but is uncomfortable because of the political ramifications? In each of these three situations, opening up investigations would imply the real possibility of prosecuting 35

Office of the Prosecutor, 2016 Report (note 25), para. 107. See for a similarly non-informative conclusion para. 230 of the same report in relation to the situation in Afghanistan. See also para. 145 of the same report in respect of the situation in Palestine. 36

Ibid., para. 230 in relation to the situation in Afghanistan, and para. 145 in relation to the situation in Palestine.

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

54

nationals, and officials in particular, of very powerful States and their allies – the United States (US) in the situation in Afghanistan, the United Kingdom (UK) in the situation in Iraq, and Israel (the US’s closest ally) in the situation in Palestine. Could this be the reason – I believe it is – that the Office of the Prosecutor hesitates in opening up investigations? I pause here to mention that not only do I believe that there is a reluctance on the part of the Office of the Prosecutor to open investigations which may lead to the prosecutions of nationals of the more powerful States, I also fully understand (though not necessarily agree with) such reluctance. It should be understood that any decision to prosecute nationals of the US, UK, and/or Israel without the blessings of the powerful States (the US and the UK), would spell catastrophe for future relations between the ICC and these States. In a fictional account of the ICC by the current author, ‘Blood in the Sand of the Justice’, the fictional Registrar observed that the contribution of a State like the US to the operations of the ICC, including in the form of intelligence, is far greater than the contribution of States parties.37 But more than that, any decision to prosecute nationals of the US or the UK, without the blessings of these States, would be the death knell for UNSC referrals. In this sense, Werle and Vormbaum’s conclusion in their article, that ‘targeting’ is not racially motivated is probably correct. What Werle and Vormbaum do not admit to, however, is that the system protects the powerful, and therefore ‘targets’ those without power, including Africans. The conclusion from above is that the reason for the disproportionate number of African cases is not, as argued by the ICC and its interlocutors, jurisdictional constraints but, at least in part, a reluctance to prosecute nationals of powerful States. But does this mean that the ICC is the villain and the AU the hero; that the ICC is the demon and the AU the angel? The answer to that question must be no. First, it would not be true to suggest that the only reason for the ‘focus’ on Africa is the geopolitical considerations. There are, without question, more conflicts, more violations, and more victims on the African continent than anywhere else in the world, suggesting that the ICC should be seized with more situations on the African continent than anywhere else.38 Second, the label of hero or angel would not, in these circumstances, 37 38

Dire Tladi, Blood in the Sand of Justice (2014), 326–327.

Shamila Batohi, Africa and the International Criminal Court: A Prosecutor’s Perspective, in: Werle/Fernandez/Vormbaum (eds.) (note 4), 50. While the conflict in Syria might affect the continued validity of this statement, there the jurisdictional constraint argument advanced by the ICC is valid.

THE ICC-AU TENSION REVISITED

55

suit the AU well. After all, the AU’s complaints of ‘targeting’ only began in 2009. Yet before 2009 – before the indictment and arrest warrant against Al Bashir – the AU never complained that the ICC only focused on Africans. It is only when the ICC began indicting heads of State – Al Bashir, Kenyatta, and Gaddafi – that the geopolitical concerns about the targeting of Africans emerge.39 Thus the AU raises concern about the ICC ‘targeting’ Africans when indictments “touch [the] African political leaders in power or their allies, but not when such indictments are against their political opponents (or rebels)”.40 Thus neither the AU nor the ICC can claim the label of angels, nor can either characterise the other as the demon when it comes to the question of ‘targeting’ of Africans. Both appear to act in the furtherance of power politics. For the ICC it is geopolitical power, while for the AU it is domestic power in each individual State that is the driving force of the respective positions.

III. Immunity Debate A. General

The immunity debate is perhaps the most visible debate in the ICC-AU tension at the moment. The intention in this section is not to reproduce the whole debate. Instead, the objective in this section is limited to an illustration of how the immunity debate itself reflects the flaws of the hero-villain trend.41 One of the most important tools that is used by actors to promote the hero-villain trend is to underplay, and often simply ignore, nuances. In the ‘targeting’ debate for example, the AU and its 39

Sosteness Francis Materu, A Strained Relationship: Reflections on the African Union’s Stand Towards the International Criminal Court from the Kenyan Experience, in: Werle/Fernandez/Vormbaum (eds.) (note 4), 212, noting that prior to the indictments of heads of State, “the ICC had already started exercising jurisdiction over cases from other referrals […] By and large, it is the indictment of African politicians who are still in power which has influenced the African Union to adopt” a hostile posture towards the ICC. 40 41

Ibid., 213.

For a comprehensive view of the author’s view on the immunity question see the following: Dire Tladi, The ICC Decisions on Chad and Malawi: On Cooperation, Immunities and Article 98, JICJ 11 (2013), 199; id., Immunity in the Era of ‘Criminalisation’: The African Union, the ICC and International Law, Japanese Yearbook of International Law 58 (2015), 17.

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

56

interlocutors ignore that the AU’s concerns are aimed principally at protecting African leaders from the reach of the ICC while other Africans appear to be fair game. On the other hand, the ICC’s defence that it is jurisdictionally constrained and that Africans have themselves referred situations conceals the real facts, namely that the ICC has avoided initiating investigations in situations involving powerful States. In the context of immunities, both sides tend to simplify the legal rules to sweep the nuances under the proverbial carpet. Time and space do not permit a detailed analysis of the various arguments. The purpose of this section is thus to mainly to consider the external motivations for the positions adopted. It is, however, necessary to recall that the debate concerning immunities is centred on two provisions of the Rome Statute. Article 27 provides that immunities do not apply before the ICC.42 Article 98 of the same Statute provides an exception from the duty of States parties to cooperate with the ICC in cases where an accused person (from a non-State party) has immunities under international law.43 I will begin by setting out the basic positions of both sides before describing the flaws of the respective positions. For the most part, both sides have focused their battle of narratives on specific situations, for example that of Al Bashir, without, it seems, considering the wider ramification of the positions they adopt. The analysis below will thus begin by a description of the respective positions before isolating the particular flaws in order to reveal the limits of the hero-villain trend. In particular, the analysis below illustrates that both sides shift their position with ease in order to achieve particular objectives, so that their respective positions can hardly be described as principled.

B. The AU’s Position Concerning Immunities of Heads of State

It is useful to begin with the position of the AU and its interlocutors on the question of immunities and Al Bashir. The position of the AU concerning the immunities 42 Art. 27 (2) Rome Statute: “Immunities or special procedural rules which may attach to the official capacity of a person, whether under national or international law, shall not bar the Court from exercising its jurisdiction over such a person”. 43

Art. 98 (1) Rome Statute: “The Court may not proceed with a request for surrender or assistance which would require the requested State to act inconsistently with its obligations under international law with respect to the State or diplomatic immunity of a person or property of a third State, unless the Court can first obtain the cooperation of that third State for the waiver of the immunity”.

THE ICC-AU TENSION REVISITED

57

of Al Bashir is based on the rules of interpretation of treaties and, in particular, on the interpretation of the Rome Statute. In 2009 the AU adopted the now famous decision in which it purported to prevent States from cooperating with the Court in the arrest and surrender of Al Bashir.44 The non-cooperation decision, which has since been repeated many times,45 sometimes with minor adjustments,46 provided, in part, that “in view of the fact that the request by the African Union has never been acted upon, the AU Member States shall not cooperate […] for [sic] the arrest and surrender of President Omar El Bashir”. As a matter of law, the decision is based on Article 98 Rome Statute.47 Yet Article 98 speaks only of “State or diplomatic immunity”, which are, in my view, specific types of immunity covering the State as such and diplomats.48 In fairness to the AU, however, the narrow reading of Article 98 which would exclude heads of State from the scope of the exception in Article 98 has not received much traction even from the ICC and its interlocutors.49 It is thus perhaps understandable that the AU does not address the potentially restricted meaning of Article 98. The AU’s position on non-cooperation is purportedly based on the principles of international law as accepted in the Rome Statute, in particular Article 98. Thus, the AU does not, in its numerous decisions on non-cooperation, appear to be rejecting or contesting the rules of the Rome Statute. Rather, it seems to be saying that the ICC is not being faithful to its own instrument by ignoring the implications of Article 98. 44 See AU Assembly, Decision on the Meeting of African States to the Rome Statute of the International Criminal Court (ICC), AU Doc. Assembly/AU/Dec.245 (XIII), 3 July 2009, para. 10. 45

See, for example, id., Decision on the Progress Report of the Commission on the Implementation of Decision Assembly/AU/Dec.270(VIV) on the Second Ministerial Meeting on the Rome Statute of the International Criminal Court (ICC), AU Doc. Assembly/AU/Dec.296(XV), 27 July 2010, para 5; id., Decision on the Implementation of the Decisions on the International Criminal Court (ICC), AU Doc. Assembly/AU/Dec.334 (XVI), 31 January 2011, para. 5; id., Decision on the International Criminal Court, AU Doc. Assembly/AU/Dec.616(XXVII), 18 July 2016, para. 2 (iv). 46 Perhaps the most significant adjustment was the insertion of a new paragraph in 2010 requesting AU Member States to balance their obligations under the Rome Statute with their obligations to the AU. See id., Decision on the Progress Report (note 45), para. 6. 47

Id. (note 44), para. 10 provides that AU Members “shall not cooperate pursuant to the provisions of Article 98”. 48

This narrow reading of Art. 98 Rome Statute is presented more comprehensively in Tladi, ICC Decisions on Chad and Malawi (note 41), 213–218. 49 For opposing views see, e.g., Claus Kreß, The International Criminal Court and Immunities under International Law for States Not Party to the Court’s State, in: Morten Bergsmo/Ling Yan (eds.), State Sovereignty and International Criminal Law (2012), 236.

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

58

Its position appears to be based on an acceptance of the application of Article 27 which provides for no immunity of an accused before the ICC with the caveat that Article 27 “applies only in the relationship between the Court and the suspect”.50 In the AU’s view – a view I share – the relationship between the Court and States (and one might add States inter se) is governed by Article 98. Yet, the AU’s apparently principled position based on an interpretation of the Rome Statute begins to shift in response to newly emerging challenges. The AU’s position with respect to Al Bashir could not be applied to the crisis concerning the indictment of Kenyatta, the head of State of Kenya. Kenya, as a State party to the Rome Statute, was cooperating with the ICC, and Kenyatta, as its head of State, could not claim immunity before the ICC. The issue of the tension between Articles 27 and 98 simply did not arise in the Kenyatta case. Kenyatta, supported by the AU, however, did not want to appear before the ICC and wanted the charges against him dropped. Suddenly, in the face of this new reality, the AU’s position shifts and is no longer based on an apparent acceptance of the validity of Article 27. The position of the AU, in response to the trials of Kenyatta and his deputy, Ruto, now appears to be based on an assertion of immunity before the ICC. In its decision at the Extraordinary Summit of 2013, the AU reaffirmed “the principles deriving from […] international customary law by which sitting Heads of State and other senior officials are granted immunity during their tenure of office.”51 On this basis the AU decides that “no charges shall be commenced or continued before any International Court or Tribunal against any AU Head of State or anybody entitled to act in that capacity during their term of office.”52 In particular, the decision states that the trials of Kenyatta and Ruto “should be suspended until they complete their terms of office.”53

50 See AU, Press Release No. 002/2012 on the Decision of Pre-Trial Chamber I of the International Criminal Court (ICC) Pursuant to Article 87(7) of the Rome Statute on the Alleged Failure by the Republic of Chad and the Republic of Malawi to Comply with the Non-Cooperation Requests Issued by the Court with Respect to the Arrest and Surrender of President Omar Hassan Al Bashir of the Republic of Sudan, 9 January 2012, 2. 51 AU Assembly, Decision on the Relationship with the International Criminal Court (ICC), AU Doc. Ext/Assembly/AU/Dec.1 (12 October 2013), para. 9. 52

Ibid., para. 10 (i).

53

Ibid., para. 10 (ii).

THE ICC-AU TENSION REVISITED

59

In addition to calling for the end of the Ruto and Kenyatta trials, the AU adopted other decisions to delegitimise Article 27 Rome Statute and thus assert the existence of immunity of (African) heads of State from the jurisdiction of international courts. First, the AU has called for the amendment of the Rome Statute to recognise immunity of heads of State before the ICC.54 Pursuant to this decision, Kenya submitted an amendment proposal, which would retain Article 27 in its current form but would add a paragraph providing that notwithstanding Article 27, “no criminal proceedings shall be instituted or continued against a sitting Head of State or Government during his or her term of office.”55 Second, the AU adopted a Protocol amending a previous Protocol (not yet in force), which would expand the jurisdiction of the African Court to cover also international crimes.56 This Protocol explicitly states that the AU Court would respect the immunities of heads of State in the context of criminal proceedings before it.57 The position of the AU, subsequent to the Kenyatta and Ruto trials, represent an astonishing shift in position from the time when the concern of the AU was the prosecution of Al Bashir. Before the Kenyatta and Ruto trials, the AU seemed to accept that the principle of immunity does not apply to trials before the ICC, although asserting (correctly in my view) that States could not be expected to cooperate with international courts in conflict with their customary international obligations (Article 98 Rome Statute). The new position, which emerges in light of the Kenyatta prosecutions, seems to be that heads of State cannot be prosecuted at all, even by international courts. While I do not agree with the merits of this proposition, for the purposes of the current point it is sufficient to only point out that the AU has abandoned its more principled position and has adopted a position based purely on the need to protect particular individuals. Thus the AU’s position appears less intended to protect the sanctity of the law, as it seems to suggest, and more to protect particular individuals, in particular serving heads of State. In this connection, it seems 54 Id., Decision on the Progress Report of the Commission on the Implementation of the Decisions on the International Criminal Court (ICC), AU Doc. Assembly/AU/Dec.493(XXII), para. 9. 55 See ICC Working Group on Amendments, Informal Compilation of Proposals to Amend the Rome Statute (on file with the author). 56 Protocol on the Amendments to the Protocol on the Statute of the African Court of Justice and Human Rights (note 6). 57

Art. 46Abis Protocol on the Amendments to the Protocol on the Statute of the African Court of Justice and Human Rights.

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

60

that when the problem concerns a head of a non-State party, the AU derives its position from Article 98 Rome Statute. But when a case concerns a head of a State party, where Article 98 is not at issue, the AU seemed too easily ready to abandon its principled position in favour of a position to protect the head of State ‘of the moment’. This oscillation of positions suggests that the AU’s challenge is not based on principle and that the legal arguments advanced by it are meant only as a smokescreen for more political objectives. The AU’s approach to international law on immunities, shifting positions based on the circumstances of the particular leaders indicted by the ICC, illustrates that, on the question of immunities, the AU cannot be regarded as angels and cannot claim the title of hero.58

C. The ICC’s Approach(es) to the Question of Immunity

The claim of the ICC (and its interlocutors), that its position on immunities is based on law and principle is equally flawed. An objective, dispassionate assessment of the facts will suggest that the only constant in the ICC’s position has been that Al Bashir must be arrested and surrendered. The legal reasons have changed like shifting sands, but the conclusion that he must be arrested has remained constant. During South Africa’s hearing on non-cooperation, I, as counsel for South Africa made the following statement: This case will have profound and far-reaching consequences far beyond Mr Al Bashir. It is not just about Mr Al Bashir. It is not just about the immunities of Mr Al Bashir. It is about the immunities of many other heads of State. But more than that, it is about the very integrity of our system, the very integrity of the Rome Statute. The whole world is watching. If legitimate legal arguments are ignored in order to come to a particular conclusion, this we believe will damage the integrity of the Court.59

58 Although, due to my legal position on immunities, I am often associated with ‘protecting the AU’, I have long held that the view that, whatever the merits of the AU’s legal arguments, the motives of the AU are less than pure. See Tladi (note 8), 69: “by suggesting, even if implicitly, that the ICC is a western tool against Africa, the AU decision [implies] that the claim to universality [of certain values was], in fact imperial wool being pulled over our collective African eyes […]”, when in fact it is the AU’s argument of imperialism “that attempt[s] to blind us from the truth”. 59

ICC, Transcript, Hearing under Article 87(7) of the Rome Statute on the Non-Compliance by South Africa with the Request by the Court for the Arrest and Surrender of Omar Al Bashir, 7 April 2017, 15, paras. 2–9, available at: https://www.icc-cpi.int/Transcripts/CR2017_02211.PDF (accessed on 30 August 2017).

THE ICC-AU TENSION REVISITED

61

The message in this quote seems clear: ‘We know you will find against South Africa even though the law supports South Africa’s claim.’ It illustrates the scepticism of the South African legal team (and perhaps the scepticism of the political authorities) towards the ability of the Court to be objective in the matter. It is useful to contextualise the basis of the scepticism by surveying the evolution of the Court’s position on immunities through the years. The first time the Court’s views on the relationship between Article 27 and Article 98 was expressed was in the Malawi and Chad decisions on non-cooperation.60 In two identical decisions, Pre-Trial Chamber I of the Court held that, notwithstanding Article 98, there remains a duty on States parties to arrest Al Bashir. The Court came to this conclusion principally on the basis that immunity before international courts is no longer recognised.61 It is unnecessary to give a full critique of the decision because, after severe criticism from nearly all quarters,62 the Court appears to have abandoned its reasoning in Malawi and Chad but certainly not the finding. It is sufficient merely to point out that the fundamental flaw in Pre-Trial Chamber I’s reasoning is that it conflates Article 27 and 98 and essentially ‘reads out of existence’ Article 98 Rome Statute.

60 Id., Pre-Trial Chamber I, The Prosecutor v. Omar Hassan Ahmad Al Bashir, ICC-02/05-01/09139, Decision Pursuant to Article 87(7) on the Failure of the Republic of Malawi to Comply with the Cooperation Requests Issued by the Court with Respect to the Arrest and Surrender of Omar Hassan Ahmed Al Bashir of 12 December 2011 (Malawi decision); id., The Prosecutor v. Omar Hassan Ahmad Al Bashir, ICC-02/05-01/09-140-tENG, Decision Pursuant to Article 87(7) on the Failure of the Republic of Chad to Comply with the Cooperation Requests Issued by the Court with Respect to the Arrest and Surrender of Omar Hassan Ahmed Al Bashir of 13 December 2011. 61

See id., Malawi decision (note 60), para. 18. For a full analysis see Tladi, ICC Decisions on Chad and Malawi (note 41). 62 See, e.g., William Schabas, Introductory Remarks: Annual Ben Ferencz Session, American Society of International Law Proceedings 106 (2012), 305; Alexander K. A. Greenwalt, Introductory Note to the International Criminal Court: Decisions Pursuant to Article 87(7) of the Rome Statute on the Failure by the Republic of Malawi and the Republic of Chad to Comply with the Cooperation Requests Issued by the Court with Respect to the Arrest and Surrender of Omar Hassan Ahmad Al Bashir and African Union Response, International Legal Materials 51 (2012), 393; Dapo Akande, The African Union’s Response to the ICC’s Decisions on Al Bashir’s Immunity: Will the ICJ Get Another Immunity Case, EJILTalk, 8 February 2012, available at: https://www.ejiltalk.org/the-african-unions-response-to-theiccs-decisions-on-bashirs-immunity-will-the-icj-get-another-immunity-case/ (accessed on 30 November 2017); Asad G. Kiyani, Al Bashir and the ICC: The Problem of Head of State Immunity, Chinese Journal of International Law 12 (2013), 467.

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

62

In the face of criticism, when the Court was next faced with the question of whether there was a duty on States parties to arrest and surrender Al Bashir, Pre-Trial Chamber II of the Court decided to adopt a different approach.63 The Court took the approach that, as a rule, Article 27 did not apply to officials of non-States parties and therefore that, ordinarily, Al Bashir would have immunity from the jurisdiction of the Court.64 The Court decided, however, that Paragraph 2 UNSC Resolution 1593 referring the situation in Darfur to the ICC, which obliged the government of Sudan to cooperate with the ICC, implicitly waived the immunity of Al Bashir.65 The Court’s conclusion is based on an apparent interpretation of Resolution 1593, yet it does not apply a single rule of interpretation to come to this conclusion.66 But more importantly for the purposes of the present article, the decision appears to be motivated by the predetermined goal to find that there was a duty on the part of the Democratic Republic of the Congo to arrest and surrender Al Bashir. Realising that it could not rely on the universally criticised decisions in Malawi and Chad, the Court sought an alternative approach which would maintain the idea that there was a duty to arrest and surrender Al Bashir notwithstanding customary international law on immunities and Article 98 Rome Statute. This alternative approach was the implicit waiver of Al Bashir’s immunities by Paragraph 2 UNSC Resolution 1593. The attendance of the AU Summit in Johannesburg in June 2015 by Al Bashir, and the subsequent appearance of South Africa before Trial Chamber II, presented an opportunity for the Court to clarify (or confirm) the legal basis of the duty to arrest and surrender Al Bashir. South Africa’s approach during the hearing was to illustrate to the Court why both the reasoning in Malawi and Chad and in DRC were flawed. With respect to the DRC decision, South Africa sought to show that a proper interpretation of Paragraph 2 UNSC Resolution 1593, relying on the customary international rules of interpretation, did not support the implicit waiver on the basis of which the Pre-Trial Chamber concluded that there was a duty to arrest and surren63

ICC, Pre-Trial Chamber II, The Prosecutor v. Omar Hassan Ahmad Al Bashir, ICC-02/05-01/09195, Decision on the Cooperation of the Democratic Republic of the Congo Regarding Omar Al Bashir’s Arrest and Surrender to the Court of 9 April 2014. 64

Ibid., para. 26.

65

Ibid., para. 29.

66

Ibid. For a detailed critique of the decision see Tladi, Immunity in the Era of ‘Criminalisation’ (note 41), 33 et seq.

THE ICC-AU TENSION REVISITED

63

der Al Bashir. In its oral arguments before the Court, South Africa implied that the Court’s approach in the DRC decision was simply to “extract implicit obligations without the application of any rules [of interpretation] whatsoever”, noting that if one goes through the DRC decision, there is “no single rule of interpretation” that was referred to or applied.67 It suggested that the Court’s approach in the DRC decision was not one of interpretation – which it described as the search for the objectively correct interpretation – but rather was a “process for the justification of policy preferences.”68 Having made this point, South Africa proceeded to show that each and every single means of interpretation (ordinary meaning, context, object and purpose, and the rule that interpretation should seek a meaning consistent with existing law) led to the conclusion that Paragraph 2 UNSC Resolution 1593 retained Al Bashir’s immunity from the jurisdiction of States parties.69 Faced with the systematic critique of the DRC decision, the Court in South Africa decided to adopt a new approach (again).70 Suddenly the reliance on the duty on the part of Sudan to cooperate in Paragraph 2 UNSC Resolution 1593, as establishing an implicit waiver of Al Bashir’s immunity, disappeared from the Court’s decision.71 The Court now adopts the position previously proposed by Akande,72 namely that the effect of a UNSC resolution – independent of its content – is to place Sudan in the position of a State party.73 This fiction is purportedly based on the fact that a referral by the UNSC has the effect that the whole treaty, including Article 27, applies. While this is undoubtedly true, what the conclusion ignores is that the whole treaty includes Article 98, and that Article 98 would apply to Sudan as a non-State party.

67

ICC, Transcript (note 59), 27, paras.1–5.

68

Ibid., 27, paras. 5–9.

69

Ibid., 27–36.

70

Id., Decision on Non-Compliance by South Africa (note 1).

71

Ibid., paras. 84–97.

72

See generally Dapo Akande, The Legal Nature of the Security Council Referrals to the ICC and its Impact on Al Bashir’s Immunity, JICJ 7 (2009), 333. 73 ICC, Decision on Non-Compliance by South Africa (note 1), para. 88 (“for the limited purpose of the situation in Darfur, Sudan has rights and obligations analogous to those of States Parties to the Statute”).

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

64

I have elsewhere provided a comprehensive critique of the Akande fiction,74 it is unnecessary to repeat those arguments here. It is, however, worth noting that, although the Chamber was unanimous in its conclusion that there was a duty on South Africa to arrest Al Bashir, the minority opinion of Judge Marc Perrin de Brichambaut disagreed with the reasoning of the Court in both the DRC decision and the South Africa decision. In an extensive opinion, Judge Perrin de Brichambaut provided extensive reasons why the Security Council resolution-based conclusions (both the implicit waiver argument in DRC and the non-State becoming like a State party fiction in South Africa) were unsatisfactory.75 Judge Perrin de Brichambaut, noting the weakness in the Court’s reasoning, himself decided to base his conclusion that there was a duty on South Africa to arrest Al Bashir on an altogether different legal basis, namely that both South Africa and Sudan were party to the Genocide Convention76 and that, as a consequence, they had both waived immunities of their officials in their relations inter se. This conclusion is quite simply astounding. To start off with, since South Africa had been invited to the hearing, one would have expected the Judge, if he were considering this line of reasoning, to at least pose the question of the applicability of Article IV Genocide Convention to South Africa during the hearing, in order to give it an opportunity to respond and to enable him to come to a conclusion that is wellreasoned and that takes into account the view of the respondent State. Instead, the respondent State is ambushed with a completely new argument when the decision is handed down (at least in the context of the proceedings) without any opportunity to rebut the argument. Second, it is well-known, as Judge Marc Perrin de Brichambaut concedes, that Article IV Genocide Convention has been understood as precluding the substantive defence of official capacity, and not procedural immunities.77 However, realising the weaknesses of the majority decision, as pointed out in his own opinion, he proceeds to construct his own interpretation with virtually no authority

74 See e.g. Tladi, ICC Decisions on Chad and Malawi (note 41), 210–212. See also id., Immunity in the Era of ‘Criminalisation’ (note 41), 33. 75

ICC, Decision on Non-Compliance by South Africa (note 1), Minority Opinion of Judge Marc Perrin de Brichambaut, paras. 28–45. 76 Convention on the Prevention and Punishment of the Crime of Genocide, 9 December 1948, UNTS 78, 277. 77 ICC, Decision on Non-Compliance by South Africa (note 1), Minority Opinion of Judge Marc Perrin de Brichambaut, para. 22.

THE ICC-AU TENSION REVISITED

65

that leads to the same conclusion as the majority in South Africa, Pre-Trial Chamber II in DRC, and Pre-Trial Chamber I in Malawi and Chad.78 The ICC has essentially adopted an à la carte approach to the applicability of immunities in the Rome Statute system, in which any approach in a wide selection of options is acceptable, as long as the option leads to the conclusion that Al Bashir must be arrested and surrendered. When the Malawi and Chad decisions were criticised as being incoherent and unjustified, the Court turned to an interpretation of Paragraph 2 UNSC Resolution 1593. When South Africa showed that the Court’s approach to Paragraph 2 falls foul of every rule of interpretation, it turned to the Akande fiction in which Sudan is deemed to be a State party to the Rome Statute – quite frankly an incredible fiction. Finally, Judge Marc Perrin de Brichambaut, recognising the flaws in the majority decision, turns to another ground, namely the waiver of immunity based on the Genocide Convention. ‘Anything to make it stick!’ The à la carte approach, incidentally, appears to also be an unwitting game of Russian Roulette. In its eagerness to ‘get Al Bashir’, the ICC decisions in DRC and South Africa both unduly and without any justification restrict the ordinary meaning of Article 27 Rome Statute. Both decisions proceed from the premise that, but for the UNSC Resolution, the ICC would be precluded from prosecuting a head of State of a non-State party – a position similar to that of the AU after the commencement of the Kenyatta trial. There is no basis for this conclusion whatsoever. On its terms, Article 27 applies to any accused that is before the Court without any qualification of nationality. I have previously criticised this position and will not repeat those views here.79 It is sufficient only to say that this view, now accepted by the ICC, is based, erroneously, on the idea that there exists in customary international law immunity before international criminal courts which cannot be excluded by a treaty for nonStates parties to that treaty. It also bears mentioning that, under the interpretation of Article 27 now advanced by the Court, it is unclear why the Court would even proceed with the preliminary assessments in Afghanistan and Palestine, since the persons likely to be investigated 78 79

Ibid., paras. 23–37.

Dire Tladi, The Immunity Provision in the AU Amendment Protocol: Separating the (Doctrinal) Wheat from the Normative (Chaff), JICJ 13 (2015), 3, 12–16.

66

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

and prosecuted in those situations would, under the restrictive interpretation of Article 27 advanced by the Court in the DRC and South Africa decisions, enjoy immunity from the Court. The Court thus appears to have cut off its nose to spite its face – get Al Bashir at all costs, even if doing so unduly restricts the ICC’s jurisdictional reach in other cases. But then again, perhaps this was strategic thinking on the part of the Court: adopt an interpretation of Article 27 that, by definition, precludes any investigation of US and Israeli officials.

IV. Conclusion The contestation for the soul of international law currently being waged between the ICC (and its interlocutors) and the AU is very often waged through misinformation, half-truths, and shifting of positions when already declared positions are inconvenient. The aim and objective of these strategies is the achievement of hero status and to paint the other side with the black brush of evil. In this context, we have seen the AU claiming that the ICC is targeting Africans but only doing so when ICC prosecutions are no longer aimed at the enemies of African leaders but at the African leaders themselves. We have also seen the ICC (and its interlocutors), in response to the targeting claims, raising, inaccurately, the jurisdictional constraints argument or the complexity argument, when in fact the reluctance to open up proper investigations can only be explained by a ‘reluctance to upset the powerful’. The analysis above has also illustrated how both sides have shifted their legal positions to align with convenient political positions. Thus, when the only concern for the AU was the indictment of heads of non-States parties like Al Bashir (and perhaps Gaddafi), the AU argument was based on Article 98 which ostensibly protects only officials of non-States parties. Yet, when the ICC’s net spread wider to entangle heads of State parties like Kenyatta, the AU’s position suddenly shifted to an (unwarranted) attack on Article 27 itself through the seeking of amendments to the Statute, adoption of an instrument designed to undermine Article 27, and calling for the end of prosecutions of African heads of States. The ICC, for its part, has been so determined to find that Al Bashir must be arrested whatever the legal constraints, that it has shifted from one legal basis to another, hoping that one, any one, will stick. In

THE ICC-AU TENSION REVISITED

67

the final decision (the South Africa decision), the weaknesses were exposed with a minority Judge criticising the Court’s reasoning, while he himself came up with a new, no more well-reasoned basis for the duty to arrest and surrender. The basic thesis advanced in this article, namely that in the battle for the soul of international criminal law, the idea that there are heroes and villains, angels and demons, white hats and black cloaks, is false, could have been advanced from an analysis of any number of issues, including cooperation and/or complementarity. However, through the chosen areas, namely the debate over targeting of Africans and the immunity debate, I hope to have shown that no one wears the white hat, that there are no heroes, and that no one is an angel in this complicated battle of wits. I hope to have shown, in this article, that all that we have are two sides (the AU and its supporters, on the one hand, and the ICC and its supporters on the other), with firmly (and perhaps honestly) held beliefs,80 that are willing to risk it all, in pursuit of pre-determined objectives. A superficial reading of this article will conclude that it represents a criticism of the ICC and the AU. But it does not. The AU and ICC are simply doing what they must: preserve their political futures. The criticism here is really aimed at the enablers – supporters or political masters who egg them on. As long as supporters of the ICC’s claim to legal neutrality are willing to turn a blind eye to the ICC’s baseless interpretation of international law in order to catch one man (who may, by the way, catch a bad cold and die tomorrow), and as long as supporters of the AU’s objective of independence and non-colonialism continue to turn a blind eye to the AU’s apparent insistence on protecting heads of State, the AU and the ICC will continue to play hero and villain games at the expense of the African continent.

Post-Script

Subsequent to the writing of this article, the Office of the Prosecutor announced, on 20 November 2017, that it would be seeking the authorisation of the Pre-Trial 80 In a private conversation, which I think the ICC Prosecutor will not mind me sharing, she said to me: “I take the position that I take regardless of the consequences, because I know that history will judge me kindly”. The statement is instructive both because it shows that the Prosecutor has no evil intent, but also because it signifies an almost dangerous messianic belief in the cause. I suspect the AU motives are equally honestly held and potentially as dangerously messianic.

68

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

Chamber for an investigation into Afghanistan.81 This means that there will, more than likely, be another non-African situation before the ICC by the time this article appears. More significantly, it means one of those situations with a potential to affect the powerful will be on the agenda of the ICC. It is worth, even if briefly, pondering whether this affects, at all, the analysis above, in particular as it pertains to the reluctance of the ICC to ‘tango’ with the powerful States. The short answer is that it does not affect, at all, my analysis or conclusions. First, the issues raised above concern not whether any investigation is opened but the length of time it has taken to open these investigations. As noted above, the length of time taken in preliminary analysis suggests to me that the ICC does want to open investigations in those situations but is reluctant to do so. Opening an investigation after a ten-year preliminary analysis when it has, for all those ten years, been obvious that crimes under the jurisdiction have been perpetrated, simply illustrates the reluctance. But there is a second more insidious reason why the Afghanistan investigations do not shake my belief that the ICC is ‘afraid’ to mess with the powerful. If investigations into Afghanistan are opened, who will face the justice of the ICC? Not a single US official because, under the DRC and South Africa decisions, Article 27 Rome Statute does not apply to officials of non-States parties, certainly not their heads of State. The decision to seek authorisation was taken with the full knowledge that the ICC’s current jurisprudence would not permit it to tango with the powerful US by indicting its head of State.

81 ICC, Situation in Afghanistan, ICC-02/17-7, Public redacted version of “Request for authorisation of an investigation pursuant to article 15” of 20 November 2017, available via: https://www.icccpi.int/Pages/record.aspx?docNo=ICC-02/17-7-Red (accessed on 11 December 2017).

FOCUS

INTERNATIONAL LAW AND THE DEHUMANISATION OF ACTIVITIES

“The System Only Dreams in Total Darkness”: The Future of Human Rights Law in the Light of Algorithmic Authority HELMUT PHILIPP AUST(

ABSTRACT: The increasing use of big data and machine learning algorithms raises several legal issues. Automated decision-making potentially undermines the very concept of human agency which is central to human rights law. Human agency enables a communication process between those bound by human rights and the rights-holders. To the extent that decision-making processes become fully automatic and autonomous, a form of algorithmic authority would arise. While human rights law is not silent with respect to such processes, doctrinal attempts to come to terms with this development are not very promising. Instead, a political process is required in order to establish a legal framework for the exercise of algorithmic authority. KEYWORDS: Algorithms, Big Data, Human Rights Law, Privacy, Non-Discrimination The system only dreams in total darkness. Matt Berninger and Carin Besser from the album ‘Sleep Well Beast’ by The National (2017)

I. Introduction On 22 November 2017, readers of the Frankfurter Allgemeine Zeitung would have been forgiven if they thought that the business section of their respectable broadsheet had turned into an outlet for science fiction literature. Under the heading “China plans total surveillance”, the newspaper reports the latest developments with respect to a plan to establish a social scoring system for all Chinese citizens. The Chinese Communist Party and the government want to establish a system based on big data (

Dr. iur., Professor of Law at the Freie Universität Berlin, Department of Law. I would like to thank Tom Burri, Prisca Feihle, Mehrdad Payandeh, and Alexander Silke for comments and criticism on an earlier draft as well as Erna Cassarà for research assistance. Any errors or misconceptions are mine.

72

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

and algorithms in order to attribute a specific social score to each and every citizen. Due to the co-operation with e-commerce company Alibaba, the system would have access to the data of the approximately 500 million customers. An enormous amount of activity would be tracked and considered for the establishment of the social score, ranging from the choice of transportation while commuting to work to the question of whether citizens comply with their statutory duty to visit their elderly parents regularly.1 Although it is difficult to understand how this scheme will eventually work in practice, it is clear that its projected power rests on algorithms and big data operations which draw inferences from the available data in a more or less automatic way. If implemented, this plan could herald a ground-breaking change with respect to how States and their citizens interact in the digital world. The Chinese plan is certainly a most radical vision of how technology can help to transform a society according to a pre-conceived ideal. Radical as it is, we should not assume that this is the future into which we are heading. But the Chinese plan is one vision of one possible future. Here and in other situations, lawyers tend to ask the question of whether the existing legal framework is still adequate for a new social development. The answer can always just be yes and no. Yes, to the extent that of course the existing rules cannot be pushed aside because of a factual development that the law aspires to regulate. And no, as law is never petrified but will, of course, adapt over time to respond to new social demands. Adaptation can then be processed by the legal system in various ways, either by means of interpreting existing rules or – if more fundamental change is needed – by changing the rules according to the processes provided for in a given legal system. This article will not ask whether human rights law is well-equipped to respond to plans and measures such as the Chinese scoring system. Rather, it takes a broader look at how new technological developments associated with the increased turn to the use of algorithms, big data processes, and self-learning machines challenge human rights law. In particular, this article focuses on the question of human agency as a fundamental notion for the construction of human rights law. It will then also reflect on various strategies deployed in the human rights discourse which are aimed at coping with the phenomena we are discussing in this article.

1 Hendrik Ankenbrand, China plant die totale Überwachung, Frankfurter Allgemeine Zeitung, 22 November 2017, 15, available at: http://www.faz.net/aktuell/wirtschaft/nationales-punktesystemchina-plant-die-totale-ueberwachung-15303648.html (accessed on 23 November 2017).

ALGORITHMIC AUTHORITY

73

Every generation goes through its own share of technological innovation. Such innovations call into question existing regulatory frameworks. While views differ as to what the function of law in a society is, one central aspect is that it is about the stabilisation of normative expectations.2 It is typical for law as a structurally conservative set of rules to catch up only over time. Accordingly, is there nothing new under the sun? Plus ça change...? It is at this point where serious disagreement begins. The digitalisation of almost all spheres of societal life in the 21st century could well be a technological revolution, not just another innovation. Revolutions tend to do away with existing legal frameworks; they do not just lead to an update along the existing lines. So much was foreshadowed by the famous dictum “Code is Law”3 which seemed to be an exaggeration back in the day. Today, self-learning algorithms and allegedly all-toomighty big data processes make us think again. Has the time come for a radical transformation of our legal frameworks? Are we entering a new phase of societal development in which the centrality of human agency is seriously called into question and which could constitute a true “paradigm change”?4 These are questions not just for human rights law. But they have a particular resonance in this field. For better or worse, the ideal of human rights is based on notions of human agency. Both the rights-holder and the addressee of human rights obligations are constructed on the ideal of a human person acting with the capability of triggering consequences through their conduct (be it deliberate or not). The human rights an individual holds are of course not directed against another individual – but against the State as a legal person, which is, if not a fiction, at least an abstraction.5 Or against private actors who some seek to bind to human rights law as many of the dangers for human rights in today’s world no longer just emanate from the State.6 In any case, the existence of the State has always been premised on the notion of it having organs. The same applies to corporate actors. These organs are represented by 2

Niklas Luhmann, Das Recht der Gesellschaft (1993), 137.

3

Lawrence Lessig, Code and other Laws of Cyberspace (1999), ch. 1.

4

Thomas Kuhn, The Structure of Scientific Revolutions (1962).

5

See already Heinrich Triepel, Völkerrecht und Landesrecht (1899), 120: “Der Staat ist zwar keine Fiktion, aber doch immer eine Abstraktion.”, which translates as “The State is not a fiction, but always an abstraction”. 6 See most recently Committee on Economic, Social and Cultural Rights, General Comment No. 24 (2017) on State obligations under the International Covenant on Economic, Social and Cultural Rights in the context of business activities, UN Doc. E/C.12/GC/24 (2017), paras. 14–22.

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

74

natural persons who deliberate, act, and thereby violate human rights – or take action in order to fulfil an obligation to protect. But regardless of the processes by which the State and its organs have arrived at their respective decisions, there has been a human element in the chain of decision-making. Even the cruellest human rights abuse is an abuse that has been carried out by humans. This is what makes the most egregious human rights violations so palpable – the fact that humans are capable of inflicting that particular kind of suffering on other humans.7 Someone is responsible for the abuse. Human rights law is about preventing such abuses – but also about finding ways and means to hold a State responsible. Human agency ensures that it is possible to give reasons for the conduct in question.8 This obviously does not mean that any conduct violating human rights is reasonable – but it can be explained by recourse to the considerations that triggered the conduct. The reasons may be dubious to the point where we start to question the common bonds among humans – but to some human being it made sense to act that way. Human rights violations can be explained – even if not necessarily understood. What makes particularly egregious human rights violations stand out is that we know that humans are presumed to be capable of empathy.9 And this capacity requires a sense for the possibility of suffering.10 This article argues that the challenge of automated decision-making to human rights law lies in calling into question core notions of human agency. Are we moving towards an era in which machines will exercise power over us? Will this lead to an exercise of authority in the form of a “mindless agency”?11 How can we come to terms with human rights violations which have been undertaken automatically, maybe even autonomously, by machines or caused by self-learning algorithms? This article proposes to analyse these questions in light of the concept of “algorithmic authority” (Deborah Lupton).12 This concept conveys the message that it is no longer just humans exercising power over other humans – but that processes based on algorithms, machine 7 This is also an important argument in the debate on the origins of modern human rights law, see to this extent the discussion of literary works and their effects on the perception of suffering in Lynn Hunt, Inventing Human Rights: A History (2008), 26–34. 8

See also Mireille Hildebrandt, Smart Technologies and the End(s) of Law (2015), 73.

9

Hunt (note 7), 39–40, on the processes of learning empathy from an early age onwards.

10

Hildebrandt (note 8), 125; see also on the visual production of empathy Susan Sontag, Regarding the Pain of Others (2003). 11

Hildebrandt (note 8), viii.

12

Deborah Lupton, Digital Sociology (2015), 49–50.

ALGORITHMIC AUTHORITY

75

learning, and big data mining may come to exercise some direct influence over individuals. It is, of course, an open question whether we will reach this stage and, if so, when. So far, no machines or self-learning algorithms operate in an environment in which there is no connection to a human actor in the decision-making process.13 Somewhere along the line, someone will be responsible. What we can observe today is that this chain of responsibility can become quite indirect. But as long as this chain exists, the usage of self-learning algorithms or autonomous machines is not a qualitatively new social phenomenon. Still, there is no shortage of literature – some high-minded, some aiming at the popular market – which tells us that this will all change in due course.14 Already today we can observe a wide pattern of social practices in which forms of machine learning, self-learning algorithms, or forms of big data mining are employed. This contribution will therefore first take stock of what we are talking about in this context. Autonomous decision-making processes and big data are elusive terms which cannot easily be defined, at least not with the cold precision that legal definitions usually exude. Still, it will be useful to flesh out in a little more detail what is usually understood by these concepts and how their realisation contributes to what is conceptualised here as algorithmic authority (II.). Subsequently, the article will reflect upon the human rights strategies that are usually deployed in order to cope with the exercise of algorithmic authority (III.). A last section will then provide some concluding observations (IV.).

II. The Construction of Algorithmic Authority Automated decision making can come in various forms. Such processes first presuppose a data operation. On the input side, data can be generated in various ways. It can be fed into an IT system by a human – for example in the case of the filing of a tax declaration. It can also be generated in the form of ‘convenience data’, i.e. data and 13

That said, States have begun to provide for legislation which enables the fully automatic issuing of administrative acts. On this development in Germany, see Thorsten Siegel, Automatisierung des Verwaltungsverfahrens: zugleich eine Anmerkung zu §§ 35a, 24 I 3, 41 II a VwVfG, Deutsches Verwaltungsblatt (DVBl.) 2017, 24. 14

For a particularly bleak scenario see Cathy O’Neil, Weapons of Math Destruction (2016).

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

76

metadata which are collected as a result of the internet activities of individuals. The automated decision-making then takes place at the level of operation or, if you will, the output side. Something is done with the data and this something takes place allegedly without (at least a direct) human intervention. At the very basis of automated decision-making are algorithms. In a leading text on the social implications of digitalisation, they are defined as a “sequence of computer code commands that tells a computer how to proceed through a series of instructions to arrive at a specified endpoint.”15 Big data, in turn, is at first sight meant to capture a vast expansion of the quantity of available data. It is a “generalized, imprecise term that refers to the use of large data sets in data science and predictive analysis.”16 It can encompass technologies that seek to maximise computational power and algorithmic accuracy. Yet, it can also describe the types of analyses which are undertaken. Furthermore, it feeds the assumption that the sheer amount of data which is analysed leads to “greater truth, objectivity, and accuracy.”17 Different devices which work based on algorithms and may contribute to big data analyses can find themselves connected in the so-called ‘internet of things’, potentially exchanging information amongst themselves without human intervention.18 The ‘internet of things’ is frequently discussed with respect to different devices in a household which are interconnected. For instance, your heating system might interact with the lights, your washing machine checks autonomously whether washing powder needs to be bought (and may order it automatically). However, this form of interconnectedness can also be taken to a different level.19 A key field for experimentation with technological utopias in this regard is the ‘smart city’.20 The vision of a smart city consists in the interconnection through data networks 15

Lupton (note 12), 11.

16

Kate Crawford/Jason Schultz, Big Data and Due Process: Toward a Framework to Redress Predictive Privacy Harms, Boston College Law Review 55 (2014), 93, 96. 17

Ibid.

18

Lupton (note 12), 9.

19

See also Christian Djeffal, Das Internet der Dinge und die öffentliche Verwaltung: Auf dem Weg zum automatisierten Smart Government?, DVBl. 2017, 808. 20 From the German policy debate see Bundesministerium für Umwelt, Naturschutz, Bau und Reaktorsicherheit and Bundesinstitut für Bau-, Stadt- und Raumforschung, Smart City Charter: Making Digital Trans formation at the Local Level Sustainable (2017), available via: http://www.bbsr.bund.de/BBSR/EN/Publica tions/SpecialPublication/2017/smart-city-charta-de-eng.html?nn=394982 (accessed on 23 November 2017).

ALGORITHMIC AUTHORITY

77

of various infrastructure devices. If these devices interact directly, it is assumed, the provision of public services can be made much more efficient. Interventions into the working of the system could be taken on a ‘real time’ basis. This vision also has a certain sustainability edge to it: If traffic flows are directed more efficiently, this would lead to a reduction of greenhouse gas emissions. In this connection, the city would become a space in which the physical world and the virtual world meet.21 Ecosystems of sensors would gather information from the urban environment. The “data-driven feedback loops” would “turn the city into a reflexive test-bed and workshop for connected habitation in enmeshed digital and physical space.”22 This would result in an information flow resting on three components, the first of which is instrumentation, involving an array of sensors measuring the environment. Second, analytics consisting of algorithms trawl the available data in order to identify patterns and predict future behaviour. Finally, so-called “actuators” would provide for “digitally controlled devices that can respond to data in real time and impact physical space.”23 Consequently, what we see would amount to a merger between social reality and digital technology,24 and, one is tempted to add, this need not be the end. In the future, digitally integrated transplants into humans could take this even further, thereby tearing down boundaries between human agency and machine-determined processes.25 Across the whole board of these concepts, a number of cross-cutting issues need to be flagged. This relates, first of all, to the notion of the ‘automated’ decision-making process. The processing of data by IT systems is, to a certain extent, always ‘automated’ once a certain command has been entered into an IT system. So much emerges already from the 1983 Volkszählungsurteil of the German Federal Constitutional Court. This decision clearly referred to the automatic character of the assessment of data under modern-day conditions.26 The need to update the protection guaranteed under the so-called allgemeines Persönlichkeitsrecht (general right of personality) into

21

Carlo Ratti/Matthew Claudel, The City of Tomorrow: Sensors, Networks, Hackers, and the Future of Urban Life (2016), 20. 22

Ibid., 23.

23

Ibid., 29.

24

Steffen Mau, Das metrische Wir: Über die Quantifizierung des Sozialen (2017), 41.

25

Ratti/Claudel (note 21), 68.

26

Federal Constitutional Court (Bundesverfassungsgericht), BVerfGE 65, 1, 42 (Volkszählung).

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

78

something more specific (clumsily translated as “right of informational self-determination”) arose, according to the Court, primarily from the absence of traditional forms of handling data and files. These would no longer be worked upon manually but automatically. In particular, if different databases were to be combined, a precise image of the personality of a given individual could be developed. The big data discourse also blurs the distinction between the quantitative and the qualitative. It is precisely the expectation that qualitatively a new set of insights will be generated following the usage of big data techniques. This qualitative turn is thus to be achieved by quantitative means. In order to generate new insights, also the ways in which data is analysed also need to change. Here, we approach the question of selflearning algorithms and actual decisions being taken without direct human involvement – a realm of fantasy as well as of fear. The idea that algorithms learn from the data they process is as fascinating as it is unsettling. It seems to open up new horizons of governing more efficiently (if we think about the usage of such algorithms in the public sector), yet raises the spectre that one day us humans could be outshone by an algorithm so smart that it is no longer controllable.27 Whereas the latter fear will – hopefully – remain within the confines of science fiction literature, the work algorithms do and the accompanying discussion take on what has been called an “algorithmic drama” in two distinctive acts: “The first act introduces algorithms as powerful and consequential actors in a wide variety of domains. […] The second act then picks up on the difficulties involved in explicating ‘how they exercise their power and influence’.”28 It is the second act that should intrigue lawyers most – how do algorithms exert “algorithmic authority”29 and how can this authority be constituted, yet also constrained by law? The present article adopts this terminology of “algorithmic authority”. This concept captures the question of the potentially changing agency engendered by algorithms, machine learning processes, and big data operations. If we are indeed moving 27 See further Thomas Burri, The Politics of Robot Autonomy, European Journal of Risk Regulation 7 (2) (2016), 341, who discusses the tensions and frictions between current operators of robots and observers of this development; see also Wolfgang Hoffmann-Riem, Verhaltenssteuerung durch Algorithmen: Eine Herausforderung durch das Recht, Archiv des öffentlichen Rechts (AöR) 142 (2017), 1, 30. 28

Malte Ziewitz, Governing Algorithms: Myth, Mess, and Methods, Science, Technology, & Human Values 41 (2016), 3, 5–6 (footnote omitted). 29

Lupton (note 12), 49–50.

ALGORITHMIC AUTHORITY

79

towards automated decision-making, it would be apt to speak of “algorithmic authority”. This authority would need to be exercised according to the law again. At least this would be the case if and to the extent that this algorithmic authority is attributable to the State. Just as States cannot escape their human rights obligations by joining international organisations or through outsourcing the provision of public goods to the private sector, nor can they escape the ties of human rights law by shifting decision-making processes to the realm of algorithms. All that is not to say that we are already living in a world of “algorithmic authority”. In this context, it is important to insist on the human factor behind the construction of algorithmic authority. The power of algorithms consists of attributing a meaning to certain input factors through specific modes of selectivity, processing, and the reduction of complexity. In this regard, they represent a specific power of definition.30 They reproduce what is considered to be valuable or relevant in a given society.31 An open question today is whether and, if so, when and how this contingency of algorithms, machine learning, and big data operations on socially constructed values might begin to disappear. Mireille Hildebrandt draws a distinction in this regard between autonomous and autonomic agents. In a nutshell, she distinguishes between autonomous behaviour which presupposes an agent capable of reflecting on her behaviour and autonomic behaviour where computing systems “with agent characteristics” have become “capable of context-dependent, nondeterministic and adaptive behaviour.” Yet, autonomic behaviour in that sense would not have reached the threshold to a fully autonomous conduct in the sense of self-reflection on part of the agent.32 Does this solve our issue with respect to the possible implications that algorithmic authority might have for human agency in general and for the requirement to make State conduct explainable in particular? I would argue that we are living in an intermediate phase of technological development in which we still do not see this form of autonomous action in the sense of Hildebrandt materialising. Instead, we see the shades of grey in-between. Decisions taken by humans rely on algorithms, machine learning processes, and big data operations in various ways. At times, these human decisionmakers themselves do not understand the rationale behind the information they get 30

Mau (note 24), 204.

31

Ibid., 205.

32

Hildebrandt (note 8), 55.

80

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

“out of the system” – the system only dreams in total darkness. All they can do is to point to the supposed rationality of the decisions, which takes us back to the assumption of a technological objectivity which is hard to sustain. So, although many of the realistic scenarios for the years to come will involve rather “autonomic behaviour” in the sense of Hildebrandt, this is no reason not to think through the implications that this will have for decision-making processes and their assessment in terms of human rights law through. On a continuum of autonomy, the end point would be autonomous behaviour where machines have become capable of reflecting on their conduct and would take decisions without any human involvement. This scenario still seems to lie in the future. What we already have are varying degrees of immediacy with respect to the impact of autonomic behaviour which involves different forms of algorithms and big data operations. It is conceivable that algorithms have informed a decision that a public official takes. The algorithms can be one factor among many or they can provide the entire background for the decision in question. In either scenario, we still have a decision taken by a public official, yet one which is based on processes which the public official may not be in a situation to fully understand. What is also already conceivable today is that certain public decisions are taken by a computer system without any direct involvement of a State organ as decision-maker. However, these decisions are still taken on the basis of an algorithm that was socially constructed. Hence, present scenarios include a weakening of the ‘human touch’ of State decisions, but it would be premature to see human agency called into question entirely. The more central algorithms, big data operations, and machine learning become for the exercise of public authority, the closer we move towards a world in which it is possible to speak of “algorithmic authority” in a meaningful sense.

III. Human Rights Strategies to Cope With Algorithmic Authority A. Doctrinal Dead Ends

Let us now turn more specifically to the level of human rights law. This article does not engage in an in-depth assessment of specific forms of the exercise of “algorithmic

ALGORITHMIC AUTHORITY

81

authority” in the light of individual provisions of human rights law. For this purpose, specific operations of these concepts would need to be identified. Yet, some of these concepts are still possibilities rather than reality. It is possible, however, to identify and assess human rights-based strategies which are deployed in our context. There is a wide variety of such strategies and they touch upon various issue areas. Those discussed here have been chosen as they represent different hopes and worries related to the technical and social phenomena we have addressed so far. However, they also illustrate – if seen together – the doctrinal dead ends of the debate. These dead ends are created by the multiple potential usages of algorithms, machine learning processes, and big data operations. It is important at this stage to refute the popular myth surrounding new technologies that they are neutral.33 Technology can be understood as a material emanation of thought processes, created for certain ends. Accordingly, it is too easy to state that algorithms, machine-learning processes, and big data operations are only particular forms of technology which can be put to beneficial uses just as well as to detrimental ones. Rather, we need to ask what the specific qualities of the data-driven approaches are and how they relate to the requirements of human rights law. And here we can see how algorithmic authority appears in different emanations. Algorithms allow for the smooth operation of systems of mass surveillance as unearthed by the Snowden revelations in 2013.34 But at the same time, carefully constructed algorithms may allow for a form of surveillance which is precisely in accordance with the requirements of human rights law. After all, it is technically conceivable to program an algorithm to the extent that certain factors – such as a person’s skin colour or ethnic background – must not be taken into account when the algorithm is in operation.35 It is another question whether this conforms to the actual practice of how such algorithms and big data mining operations are used in the intelligence and security communities. But theoretically, the technological advances are just as threatening for notions of privacy and informational self-determination as they could act as guardians of these notions against an overreaching State.

33

See also ibid., 162–163.

34

On this development see Helmut Philipp Aust, Spionage im Zeitalter von Big Data: Globale Überwachung und der Schutz der Privatsphäre im Völkerrecht, Archiv des Völkerrechts 52 (2014), 375. 35

See Timo Rademacher, Predictive Policing im deutschen Polizeirecht, AöR 142 (2017), 366, 398.

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

82

Beyond this debate, there is also room for considerable disagreement on the contour of the right to privacy in the digital age.36 The United Nations (UN) General Assembly has affirmed at various times that this human right is also to be protected in the online world.37 In fact, it has been affirmed that human rights need to be complied with online just like offline.38 Beneath these by now well-worn formulae, considerable uncertainty hovers which concerns already the identification of what is exactly understood by privacy. In this respect, a transatlantic divide has been identified with respect to a schism between a strict position on privacy in Europe (allegedly going back to the case law of the German Federal Constitutional Court) and a more pragmatic position on the other side of the Atlantic where only reasonable expectations of privacy are protected.39 In other words, if data is shared with others, it loses any protection of a right to privacy under United States constitutional law. Arguably, these different understandings of privacy in the constitutional realm will also have an impact on the contours of the human right to privacy under international human rights agreements.40 At least the domestic law understandings of privacy provide a bedrock upon which a common understanding of privacy on the international level will need to be built. And there is more to the disagreement about privacy than just this perceived transatlantic divide. How to integrate Japanese understandings of animism (which more easily attribute a soul to things) into our understanding of privacy, for instance?41 And how does a lack of privacy in the rights culture in authoritarian States impact on the debate? What looks like a human right set in stone from a European perspective and what 36 Human Rights Council, Report of the Special Rapporteur on the Right to Privacy, Joseph A. Cantanacci, UN Doc. A/HRC/31/64 (2016), para. 20. 37

United Nations General Assembly (GA), The Right to Privacy in the Digital Age, UN Docs. A/ RES/68/167 (2013); A/RES/69/166 (2014); A/RES/71/199 (2016). 38

Most recently in id., UN Doc. A/RES/71/199 (2016), op. para. 3.

39

James Q. Whitman, The Two Western Cultures of Privacy: Dignity versus Liberty, Yale Law Journal 113 (2004), 1151, 1155; Hildebrandt (note 8), 83; Russell A. Miller, A Rose By Any Other Name? The Comparative Law of the NSA-Affair, in: Russell A. Miller (ed.), Privacy and Power: A Transatlantic Dialogue in the Shadow of the NSA-Affair (2017), 63, 79; see also Karl-Ferdinand Gärditz/CarlFriedrich Stuckenberg, Vorratsdatenspeicherung à la américaine?, Juristenzeitung 69 (2014), 209. 40 In this respect, the analytical category of the so-called “consubstantial norms” is useful, as these norms have a basis in both international and domestic law, yet overlap largely in terms of content, see further Antonios Tzanakopoulos, Judicial Dialogue as a Means of Interpretation, in: Helmut Philipp Aust/Georg Nolte (eds.), The Interpretation of International Law by Domestic Courts: Uniformity, Diversity, Convergence (2016), 72, 86–88. 41

See further on this Hildebrandt (note 8), 122–130.

ALGORITHMIC AUTHORITY

83

seems to be universally endorsed by all UN Member States through UN General Assembly resolutions unravels before our eyes if we look more carefully and with a willingness to read between the lines – especially as the UN General Assembly resolutions are full of constructive ambiguity in their formulations. If one looks at human rights law to keep in check the exercise of algorithmic authority, it might be more sensible to turn to the principle of non-discrimination.42 Here, it should be easier to find common ground with respect to the results of the exercise of algorithmic authority. Processing data through sets of algorithms, machine learning processes, and big data operations must not lead to discrimination based on suspicious criteria such as gender, race, or religion. And such forms of discrimination might enter the picture here with a twist. Recourse to big data analytics has at times been motivated so as to avoid precisely the forms of discrimination just described. Racial profiling is a case in point and might be replaced with a sophisticated algorithm indicating to the police officer who is to be controlled in the public sphere. But what if the data operation replicates the discrimination through a process which is based eventually on a socially constructed set of predictors? In this respect, it needs to be asked whether it will be possible to make algorithms, machine-learning processes, and big data operations ‘unlearn’ certain operations so as not to re-introduce suspicious criteria through the backdoor, then justified on allegedly positivist and objective data operations.43 ‘Unlearning’ may not always be the correct technical expression – but it conveys the meaning that it is necessary to control the operation of algorithms ex post so that they do not yield results arrived at by unlawful means (this could also be called ‘debiasing’). Too effective an unlearning process might then, however, considerably reduce the attractiveness of the processes underlying algorithmic authority. After all, it is precisely the capability of data operations to rely on all available data which makes them so appealing. In particular, the promise is to generate insights from a systematic analysis of data which will also reveal the unexpected. In any case, it is a difficult question to determine where the boundary between reliance on a forbidden predictor and its mere incidental use will come down. If ‘race’ was just one of the factors which were somehow processed, does this render inadmissi-

42

On this see also ibid., 94–96.

43

Ibid., 193.

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

84

ble any kind of reliance on results which can also be explained, retroactively, by a number of other factors? In other contexts, the turn to algorithmic authority may be perceived as completely beneficial in the first place. Consider, for instance, the role that algorithms, machine learning, and big data operations might play in the context of the realisation of the Sustainable Development Goals (SDGs) as adopted by the UN General Assembly in 2015.44 While it is a matter of broader debate to what extent the SDGs are based on ideals of human rights or whether human rights as a meaningful category have been largely side-lined in the document,45 it is indisputable that many of the concrete goals of the SDGs have a close connection to the realisation of social and economic rights. In this context, big data can help State agencies to provide better public services, some of which may be called for by positive obligations under international human rights law. Recourse to forms of predictive policing can help ensure access to safe public spaces as required by SDG 11.46 Or think of a better allocation of fresh water resources, thereby helping to implement a right to water.47 But here again threats and challenges are lurking. For a variety of reasons, the public sector will be hard-pressed to implement many of these sensible approaches on its own – especially in States in the so-called ‘Global South’.48 Rather, the public sector will have to rely on cooperation with the private sector, often in the form of publicprivate partnerships (PPPs). While PPPs are not objectionable as such, recourse to them in the context of the highly complex data operations we are discussing here comes with a price attached. Frequently, private companies will not want to disclose 44 Which is also acknowledged in the 2016 resolution of the GA on the right to privacy in the digital age, see UN Doc. A/RES/71/199 (2016), preamble, op. para. 2. 45 For a critical view see Carmel Williams/Paul Hunt, Neglecting Human Rights: Accountability, Data and Sustainable Development Goal 3, The International Journal of Human Rights 21 (2017), 1114. 46

On SDG 11 see further Helmut Philipp Aust/Anél du Plessis, Good Urban Governance as a Global Aspiration: On the Potential and Limits of Sustainable Development Goal 11, in: Duncan French/Louis Kotzé (eds.), Sustainable Development Goals: Law, Theory and Implementation (2018), forthcoming. 47 For a policy brief on the use of big data in this regard see Independent Expert Advisory Group on a Data Revolution for Sustainable Development, A World that Counts: Mobilising the Data Revolution for Sustainable Development (2014), available via: http://www.undatarevolution.org/report/ (accessed on 23 November 2017). 48

This expression is used here in the sense of a ‘concept metaphor’, see further Sophie Oldfield/Susan Parnell, From the South, in: Susan Parnell/Sophie Oldfield (eds.), The Routledge Handbook on Cities of the Global South (2014), 1, 2; Ananya Roy, Worlding the South: Towards a Post-Colonial Urban Theory, in: ibid., 9, 15.

ALGORITHMIC AUTHORITY

85

the concrete workings of algorithms or big data operations for the fear of losing a comparative advantage over other competitors. Private sector companies might seek refuge in patent law, thereby creating regime conflicts between human rights-based needs of disclosing the workings of algorithms and the protection of secrecy for the sake of business interests. Arguably, the public sector cedes part of its autonomy in decision-making processes to the private sector. This will also entail a complication for accountability and oversight processes.49 It is here where the central tenet of human agency of human rights law is most under pressure. While there might not be a universal right to know why State action has been taken, it is an essential element of both the rule of law and modern understandings of human rights law that exercises of State power must be justified. This is also an important requirement for the concept of the separation of powers, especially with a view to enabling courts to meaningfully control the exercise of public power. Based on the information conveyed about State action, it becomes possible to argue about its relation to human rights law.50 This can become difficult when the public sector relies on powerful algorithms and big data operations, especially when this form of conduct involves cooperation with the private sector. The provision of public goods becomes privatised in this form even when the State is still taking the relevant decisions. Algorithmic operations are, to a certain extent, arcane in nature – with all the consequences attached with respect to transparency51 – the system only dreams in total darkness. At the same time, one should not be too alarmist in this regard. Today, it seems that decisions impacting negatively on the freedom of individuals are rarely taken based solely on automated decision-making. This is especially so when it comes to infringements of the traditional civil and political rights by direct State action. But the case of the use of algorithms, machine learning processes, and big data operations for the fulfilment of the SDGs alerts us to the questions of distributive justice where the impact of “algorithmic authority” will in all likelihood soon be felt more acutely. While there was for a long time a tendency in human rights scholarship – especially in the West –

49 For a thorough discussion of the implications of privacy on the notion of public spaces and related questions of accountability see Daniel Moeckli, Exclusion from Public Space: A Comparative Constitutional Analysis (2016), 53–61. 50

Hildebrandt (note 8), xi.

51

Mau (note 24), 206.

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

86

to take economic and social rights less seriously, this appears to be changing.52 The realisation of the SDGs will require the allocation of resources – and with this allocation conflicts between the realisation of different social and economic rights are not implausible. This reminds us that every human right needs to be framed in the light of other human rights.53 How will algorithms and machine-learning processes weigh different human rights considerations? Will it be possible to program algorithms in a way that they can factor in the principle of “systemic integration”, as the International Law Commission has labelled Article 31 (3)(c) Vienna Convention on the Law of Treaties?54

B. Escape Routes

There are two usual escape routes out of the dilemma of how to preserve ‘human sovereignty’ over the data-driven processes with which we are dealing here: a turn to regulation and an emphasis on transparency. These escape routes amount to what could be described as second order arguments about algorithmic authority and human rights. They are based on the assumption that it will be futile to hedge in the excessive consumption of data which stands in contrast to established notions of privacy. Instead, the focus should shift to either regulation of how data is used or at least on ensuring a measure of transparency about the autonomous processes on which decisions are eventually based. It has been suggested early on in the debate about big data processes that what is needed is a maxim of “procedural data due process” which would “regulate the fairness of Big Data’s analytical processes.”55 To a certain extent this amounts to a capitulation in the face of social reality. Irrespective of whether this is a convincing move, it is also far from guaranteed that the turn to regulation will be able to impact effectively on the 52 For a passionate call for more attention to social and economic rights see Philip Alston, The Populist Challenge to Human Rights, Journal of Human Rights Practice 9 (2017), 1, 9–10. 53

Effy Vayena/John Tasioulas, The Dynamics of Big Data and Human Rights: The Case of Scientific Research, Philosophical Transactions of the Royal Society A 374 (2016), 6; see also Hildebrandt (note 8), xiii. 54

International Law Commission, Fragmentation of International Law: Difficulties Arising from the Diversification and Expansion of International Law, Report of the Study Group of the International Law Commission, finalized by Martti Koskenniemi, UN Doc. A/CN.4/L.682 (2006), para. 413. Vienna Convention on the Law of Treaties, 13 May 1969, UNTS 1155, 331. 55

Crawford/Schultz (note 16), 109.

ALGORITHMIC AUTHORITY

87

social reality against which it has been conjured up. As mentioned before, algorithmic authority is often derived from data operations undertaken on the basis of software provided by IT companies with a global reach. These companies operate smoothly through a global web of regulation. Effectively regulating the use of data by these companies presupposes a certain jurisdictional resolve, especially the willingness to enforce these rules also vis-à-vis other powerful actors from both the private sector and among States. This is not to say that attempts to turn to regulation are necessarily doomed. In fact, recent developments at the European Union (EU) level show how limits on the use of personal data are sought for and that this may entail serious jurisdictional struggles.56 The size of the EU market should ensure, however, that these attempts are taken seriously by relevant IT companies. Regulation is thus not necessarily a lost cause – but it depends on a robust political will to enforce European rules in a transatlantic setting. The turn to transparency is also not without complications.57 It responds to a widely shared feeling that a loss of transparency is one of the main problems with respect to the data-driven processes analysed in this contribution.58 It would be this lack of transparency that would stand in the way of giving reasons as we are accustomed to with respect to State action triggered by human agency. After all, if it were possible to see through the processes underlying the exercise of algorithmic authority, it would be sufficient for a State organ to ratify the decision, to fully understand the rationale behind the data operation, and to translate these insights into the regular forms of communication between the State and its citizens. If it were so simple this would indeed be the best way out of the issues we are discussing in this article. Yet, it can be doubted, first, whether the processes behind the exercise of algorithmic authority lend themselves to such an identification and translation process. What 56

European Court of Justice, Maximilian Schrems v. Data Protection Commissioner, joined party: Digital Rights Ireland Ltd, Case C-362/14, Judgment of 6 October 2015; in 2018, the EU General Data Protection Regulation will enter into force, see Regulation (EU) 2016/679 of the European Parliament and European Council of 27 April 2016 on the protection of natural persons with respect to the processing of personal data and on the free movement of such data, and repealing Directive 95/46/EC (General Data Protection Regulation) of 4 May 2016, OJ 2016 L 119, 1. 57 For a thoughtful exploration of the conceptual issues surrounding transparency see Andrea Bianchi, On Power and Illusion: The Concept of Transparency in International Law, in: Andrea Bianchi/Anne Peters (eds.), Transparency in International Law (2013), 1, 6–10. 58

Fleur Johns, Global Governance Through the Pairing of List and Algorithm, Environment and Planning D: Society and Space 34 (2016), 126, 140.

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

88

to do when “the system only dreams in total darkness”? Second, it is also far from clear whether transparency is the adequate response in the first place. One person’s transparency is another one’s violation of the right to privacy.59 Especially the coupling of data sets in big data analyses might require the disclosure of data operations which affect the data of other persons. This would not necessarily be a human rights infringement if the data is just processed and then deleted – and if the person becomes not identifiable with the data which was used for the respective processes (if that is technically feasible at all). On a more conceptual level, the turn to transparency raises questions with respect to the expectations that come attached to it. As Fleur Johns has argued, we need to ask ourselves whether there is a “prospect of the world’s unconcealment”.60 She argues that the turn to transparency might stand for a wish to affirm “human authority over data in principle [rather] than […] exposing particular list-plus-algorithms to effective scrutiny.”61 Seen from this perspective, the turn to transparency is unmasked as a more or less symbolic move.62

IV. Concluding Observations What should be clear with respect to both the quantitative and the qualitative phenomena: Algorithms and big data are not “objective, complete and neutral”. Instead, their production and usage are “political, social and cultural processes.”63 In other words, there is nothing inescapable about algorithms and big data. It is very likely that they will change the world we are living in. Yet, the precise way in which this will take place will still be the subject of political decision-making processes. Of course States are not omnipotent in this regard. They will need to attempt to regulate a field in which private actors often take the lead. Yet, this makes the political discussion of what we expect ‘the State’ to do even more acute. I would second the caution59

Hildebrandt (note 8), 74.

60

Fleur Johns, Data, Detection, and the Redistribution of the Sensible, American Journal of International Law 111 (2017), 57, 84. 61

Johns (note 62), 142.

62

See also Bianchi (note 61), 15–19.

63

Lupton (note 12), 101.

ALGORITHMIC AUTHORITY

89

ary position of Johns that it may be time “to suspend, for a while, our appetite for some way for the law to resolve, decisively, the politics of list-plus-algorithms, or to bend the latter to existing doctrine.”64 The emerging debate about human rights and algorithmic authority comes at a time when the international system of human rights is under manifold pressures. Many see a backlash occurring against human rights, some speak of a crisis of both the ideals and institutions of human rights (law).65 Are we living in the “endtimes of human rights”66 or can we see at least a “twilight of human rights law”67 as some scholars suggest? These debates have been triggered by a loss of a sense of direction – it is no longer taken for granted that human rights are the right project to put the planet on a path towards more justice and well-being for all. Two considerations are relevant for the present article in this context. First, the evidence that the human rights project is doomed is by no means uncontested. Especially among scholars working with empirical data about human rights progress, views differ as to the truths that can be gleaned from the data.68 More important, however, is the second point: The backlash discourse invites us to rethink what the human rights project can offer to the digital world in which we are now living. Is it reasonable to think that human rights law will be the right place to look for an answer for the de-humanising tendencies of algorithmic authority? Here I have my doubts. This has to do with perceptions as much as with political considerations. Human rights are often conceptualised as rights for the weak and oppressed. In our digital lives, we often subjugate ourselves, however, and we do that voluntarily on the basis of our (hypothetically) free and informed consent. It cannot be expected to frame the legal regime for algorithmic authority out of life experiences of broad parts of the population who do not view their practices as triggering human rights questions. From a doctrinal perspective it is of course correct that the existence of human rights violations does not depend on a broad and shared understanding that a certain practice is a human rights violation. What is needed in order to sustain human rights law, 64

Johns (note 62), 144.

65

See Alston (note 56).

66

Stephen Hopgood, The Endtimes of Human Rights (2013).

67

Eric A. Posner, The Twilight of Human Rights Law (2014).

68

For a more positive assessment as compared to Hopgood and Posner, see Kathryn Sikkink, Evidence for Hope: Making Human Rights Work in the 21st Century (2017).

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

90

however, is a certain social substratum.69 There is widespread social activism with regard to the politics of algorithms, big data, and machine learning. In order to make an impact, these forms of activism need to translate into political majorities. If the potential rights-holders do not recognise their conduct to be covered and protected by human rights, it becomes paternalistic to impose a human rights philosophy on their dealings from a certain point onward. Much of what we worry about when we think about “algorithmic authority” is not conduct emanating from the State. Rather, it is the conduct of non-State actors, many of them powerful corporations with a global reach like Amazon, Facebook, or Google. Accordingly, we want States to regulate their conduct. This regulation presupposes a political struggle. It is naïve to think that the contents of the regulation can all be neatly deduced out of some basic considerations of human rights law. Also, the idea of positive obligations – important as it is – will not be able to deliver here on its own. States have positive obligations – but they also enjoy considerable discretion with respect to how they will implement these positive obligations. This discretion is the space for politics.

69

See, mutatis mutandis, Luhmann (note 2), 132.

International Law and Artificial Intelligence THOMAS BURRI(

ABSTRACT: This article proposes five arguments about major aspects of artificial intelligence and their implications for international law. The aspects are: automation, personhood, weapons systems, control, and standardisation. The arguments in aggregate convey an idea of where international law needs to be adapted in order to cope with the artificial intelligence revolution under way. The arguments also show the inspiration that may be drawn from existing international law for the governance of artificial intelligence. KEYWORDS: Artificial Intelligence, Robotics, Ethics, International Law, AI Personhood, Autonomous Weapons, Automation, Control, IEEE

I. Introduction This paper proposes five arguments about major aspects of artificial intelligence and their implications for international law. The aspects are: automation, personhood, weapons systems, control, and standardisation. Each section of this paper makes an argument about one aspect. Since the aspects are interlinked, in the end, hopefully, a broad picture emerges that depicts where international law and artificial intelligence stand. International lawyers have so far focused almost exclusively on autonomous weapons systems. This focus is too narrow and does not do justice to the complexity and the reach of artificial intelligence. Artificial intelligence is about much more than just weapons. Hence the underlying idea of this paper – to broaden the discussion in inter( Prof. Dr., Assistant Professor of international law and European law at the University of St. Gallen (HSG) in Switzerland; Dr. iur. (Zurich), LL.M. (College of Europe, Bruges), lic.iur. (Basel), admitted to the bar of the canton of Zurich, Venia Legendi for international law, European law, and constitutional law. Contact: , www.thomas-burri.com. My thanks go to Helmut Aust for helpful comments and the Editors of the German Yearbook of International Law for support and helpful comments.

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

92

national law on artificial intelligence. While the paper discusses the broad implications of artificial intelligence for international law, it also highlights, vice versa, what international law implies for artificial intelligence and how it can provide inspiration. The aim of this paper is not to state absolute truths. Some of the arguments the paper makes are too bold to be absolutely true. Artificial intelligence is also a new phenomenon – at least in its present form – and it is too early to state absolute truths, at least when it comes to the laws and ethics governing it and governed by it. Ideally, the paper, in its argumentative way, serves to show avenues to be explored by international lawyers and to launch a broader discussion in international law.

II. Argument 1: International Law Will Not Be Automated The work of lawyers is increasingly automated. Lawyers organise themselves and their work in new ways, relying on deep databases for research and information technology for their services. But beyond this rather basic digitalisation, which most professions have experienced in one way or another, tasks that had before required lawyers’ unfettered attention are being more and more “mechanised”1. Routine jobs such as the drafting of contracts, appeals against rulings, or due diligence scrutiny are increasingly automated.2 Moreover, machines are also beginning to outperform lawyers in legal assessment, which has traditionally been their core competence. Computer programmes, for instance, predict the outcome of appeals against tax rulings and fines for traffic offences with great accuracy.3 This evolution will likely continue and accelerate in the near future. It will reduce the role of humans in legal services and expertise, leaving lawyers with less work and fewer hours to bill. 1

“Mechanisation of the Law” is a term a research group including John Armour, Horst Eidenmüller, Sarah Green, Jeremias Prassl, Mari Sako, and Rebecca Williams at the University of Oxford uses to describe the phenomenon, available at: https://www.law.ox.ac.uk/research-and-subject-groups/researchcollection-law-and-technology/blog/2017/03/mechanisation-law (accessed on 22 November 2017). 2 For illustration see, for instance, the blog by Daniel Martin Katz, Michael Bommarito, and Jon Zelner, available at: www.computationallegalstudies.com (accessed on 22 November 2017). 3

See “Blue J Legal”, available via: https://www.bluejlegal.com (accessed on 22 November 2017). For the United States Supreme Court, see Daniel Martin Katz/Michael James Bommarito II/Josh Blackman, A General Approach for Predicting the Behaviour of the Supreme Court of the United States, 9 July 2014, available at: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2463244 (accessed on 22 November 2017).

INTERNATIONAL LAW AND ARTIFICIAL INTELLIGENCE

93

It is artificial intelligence that transforms legal work. Machine learning, the type of artificial intelligence that has evolved the most in recent times, has become most effective.4 While machine learning is commonly associated with automated driving, search algorithms, and image recognition, it extends to cover the law.5 The learning of a machine, however, crucially depends on the quantity and quality of the data it is fed. The bigger and cleaner the training data, the better the learning and the more solid the intelligence a machine learning algorithm can apply to newly gathered data. Moreover, structure in data, including categories and variables, improves machine learning. Consequently, when the law is extensive, dense, homogeneous, and structured and involves defined sets of variables, it lends itself to the application of machine learning. This is the case in some domains of the law, for instance, in national tax law. The fields to be filled in on tax forms are clearly defined. The disputes that arise in tax law tend to hinge on details and nuances. Disputes are abundant, though, and municipal tax authorities and courts each year hand down thousands of rulings which are rather clear and, moreover, often accessible in public databases. The resulting body of law is large, solid, and uniform – and thus ideal for machine learning. The same holds true for other areas of the law, such as traffic law, asylum law, or the law of mergers and acquisitions, which are all marked by rulings and routines that are as endless as they are frequent. International law is different though.6 It lacks the above quantitative and qualitative characteristics. The international legal order is relatively small and diverse. With fewer than 200 States and a few dozen international entities, it involves only a limited 4 See Thomas Burri, Machine Learning and the Law: 5 Theses, 7 March 2017, accepted at NIPS 2016 in Barcelona, forthcoming in JMLR Workshop and Conference Proceedings, available at: https:// papers.ssrn.com/sol3/papers.cfm?abstract_id=2927625 (accessed on 22 November 2017). 5

Machine learning and the law is a rather new field. However, the conference ‘Fairness, Accountability, and Transparency in Machine Learning’ has been organised on a yearly basis since 2014, available via: http://www.fatml.org (accessed on 22 November 2017). Various conference series and organisations have also covered the topic broadly for some time, e.g. CEPE (Computer Ethics: Philosophical Enquiry), ETHICOMP, and IACAP (International Association for Computing and Philosophy). NIPS (Neural Information Processing Systems) also offered a workshop on Machine Learning and the Law in 2016, organised by Adrian Weller, Jatinder Singh, Thomas D. Grant, and Conrad McDonnell, available via: www.mlandthelaw.org (accessed on 22 November 2017) (with papers). 6 Scholarship on international law and artificial intelligence is hard to find, save when concerning autonomous weapons system; on that, see argument 3, infra, IV. On law and technology more broadly see, for instance, Ugo Pagallo, The Law of Robots: Crimes, Contracts, Torts (2013); see also the Springer journal ‘Artificial Intelligence and Law’ (published since 1992).

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

94

number of actors; the number remains comparatively low, even if non-governmental organisations and transnational companies are factored in. The law is laid down in a limited number of treaties which cover the widest range of issues and are often worded in general and vague terms.7 Universally applicable multilateral treaties are the exception, rather than the rule. Customary international law is notoriously fuzzy.8 In line with the strong diplomatic tradition in international law, disputes among actors are settled amicably and in secrecy. International court cases and authoritative rulings occur only exceptionally.9 The relatively few rulings that exist cover a wild variety of issues. Where rulings on international law are more numerous, namely in municipal law, they are often inaccessible and written in various languages. The international law rulings of municipal courts are generally marked strongly by domestic law and facts – in both the domestic implementation of international law and interpretation more generally – so that they do not scale well.10 There are areas of special international law where the law is denser and more homogeneous, for example the law of the World Trade Organization (WTO) or international criminal law. But rulings are still relatively few. The WTO counted 529 disputes as per 8 September 2017.11 The International Criminal Tribunal for the 7 The truth of these characteristics is evident from any international law manual; by way of example: James Crawford, Brownlie’s Principles of Public International Law (8th ed. 2012); Thomas Buergenthal/ Sean D. Murphy, Public International Law in a Nutshell (5th ed. 2013); Jan Klabbers, International Law (2013); in German: Andreas von Arnauld, Völkerrecht (3rd ed. 2016) or Matthias Herdegen, Völkerrecht (16th ed. 2017). 8

See, again by way of example, how the International Court of Justice (ICJ) had to go to great length to discuss the customary nature of only a handful of rules in ICJ, North Sea Continental Shelf Cases, Judgment of 20 February 1969, ICJ Reports 1969, 3, and id., Military and Paramilitary Activities in and against Nicaragua (Nicaragua v. United States of America), Merits, Judgment of 27 June 1986, ICJ Reports 1986, 14. 9

This is not to say that international law has not grown over the past decades and that international court rulings have not become more numerous over the recent past. (See Karen J. Alter, The New Terrain of International Law: Court, Politics, Rights (2014).) The statement is not made to debase international law, either. It is made solely in the perspective of what is needed to put artificial intelligence and machine learning to work. 10

See the Oxford Reports on International Law, with the section International Law in Domestic Courts, which increases access to municipal court decisions, but is also a testimony to the variety of law and cases, available via: http://opil.ouplaw.com/home/ORIL (accessed on 27 November 2017); for scholarship see Helmut Philipp Aust/Georg Nolte (eds.), The Interpretation of International Law by Domestic Courts: Uniformity, Diversity, Convergence (2016). 11

See World Trade Organization, Dispute Settlement, available at: https://www.wto.org/english/ tratop_e/dispu_e/dispu_e.htm (accessed on 27 November 2017).

INTERNATIONAL LAW AND ARTIFICIAL INTELLIGENCE

95

former Yugoslavia indicted 161 persons and handed down some 100 judgments.12 Even if the considerable volume filled by statements and evidence at the Tribunal served as supplementary material to the rulings,13 the data remains too limited for artificial intelligence to learn to give a sound legal assessment. The general conclusion for international law is that neither international law nor the work of international lawyers will be automated.

III. Argument 2: As Artificially Intelligent Entities with Legal Personality Emerge, the Law Must Be Reviewed An artificially intelligent entity is a legal person governed by national law that ‘houses’ an artificial intelligence. Shawn Bayern first came up with the idea to establish a limited liability company under current United States law, i.e. the law as it stands now, and put an artificial intelligence (or, synonymously, an autonomous system, a smart algorithm, a software agent, etc.) wholly and solely in charge of it.14 Bayern explained in detail how this construction works and showed that it has the effect of bestowing legal personality on an artificial intelligence. Bayern and I, together with four co-authors, further explored whether the construction is operable under other legal systems and what it implies for constitutional law.15 Lynn LoPucki confirmed that the creation of such “algorithmic entities” was lawful.16 He argued that their emergence was unavoidable, but also most consequential. Criminals would likely use artificially intelligent entities for their purposes and profit 12

See International Criminal Tribunal for the former Yugoslavia (ICTY), Infographic: ICTY Facts & Figures, available at: http://www.icty.org/en/content/infographic-icty-facts-figures (accessed on 27 November 2017) and id., Judgment List, available at: http://www.icty.org/en/cases/judgement-list (accessed on 27 November 2017). 13

Some 2.5 million pages, see ICTY, Infographic (note 12).

14

Shawn Bayern, The Implications of Modern Business-Entity Law for the Regulation of Autonomous Systems, European Journal of Risk Regulation 2 (2016), 297, also published in Stanford Technology Law Review 19 (2015), 93. 15

Id. et al., Company Law and Autonomous Systems: A Blueprint for Lawyers, Entrepreneurs, and Regulators, Hastings Science & Technology Law Journal 9 (2) (2017), 135. 16 Lynn M. LoPucki, Algorithmic Entities, Washington University Law Review 95 (forthcoming 2017), 76, available at: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2954173 (accessed on 27 November 2017).

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

96

most from them. In a recent paper, I then showed the implications of artificially intelligent entities for the internal market of the European Union.17 It turns out that the internal market’s freedom of establishment has the effect that the legal personality of an artificially intelligent entity established pursuant to the law of a European Union Member State has to be recognised by all other Member States. In other words, mutual recognition of national legal personality, which the Court of Justice of the European Union established through the rulings in Daily Mail, Centros, and Ueberseering,18 applies to artificially intelligent entities just like to all other national legal persons (provided they pursue an economic activity). The resulting freedom of these entities to do business in the entire European Union – “the free movement of algorithms” – can only be restricted in very narrow circumstances. Permissible restrictions vary depending on whether artificially intelligent entities are controlled or uncontrolled by humans. Clearly, LoPucki is right to sound the warning bell. In the past and present, the coexistence of more than 200 competing national laws has made possible the exploitation of minimal company law rules for crime and tax avoidance. Shell companies in Panama and elsewhere are one prominent example.19 A similar effect has been witnessed with flags of convenience of ships, which allowed ship owners to avoid strict national safety standards, labour laws, and taxation.20 Similar evasion, avoidance, and abuse of law may occur with artificially intelligent entities, since establishing them is 17

Thomas Burri, Free Movement of Algorithms: Artificially Intelligent Entities Conquer the European Union’s Internal Market, in: Woodrow Barfield/Ugo Pagallo (eds.), Research Handbook on the Law of Artificial Intelligence (forthcoming 2017/2018), available at: https://papers.ssrn.com/sol3/ papers.cfm?abstract_id=3010233 (accessed on 27 November 2017). 18 European Court of Justice (ECJ), Daily Mail, Case C-81/87, Judgment of 27 September 1988, 5483; id., Centros, Case C-212/97, Judgment of 9 March 1999, and id., Ueberseering, Case C-208/00, Judgment of 5 November 2002. 19 Compare International Court of Justice, Barcelona Traction, Light and Power Company, Limited (New Application: 1962) (Belgium v. Spain), Judgment of 5 February 1970, ICJ Reports 1970, 3, para. 56: “[T]he law has recognized that the independent existence of the legal entity cannot be treated as an absolute. It is in this context that the process of ‘lifting the corporate veil’ or ‘disregarding the legal entity’ has been found justified and equitable in certain circumstances or for certain purposes. The wealth of practice already accumulated on the subject in municipal law indicates that the veil is lifted, for instance, to prevent the misuse of the privileges of legal personality, as in certain cases of fraud or malfeasance, to protect third persons such as a creditor or purchaser, or to prevent the evasion of legal requirements or of obligations.” And para. 59: “It follows that on the international plane also there may in principle be special circumstances which justify the lifting of the veil in the interest of shareholders”. 20

See the famous case ECJ, Viking, Case C-438/05, Judgment of 11 September 2007.

INTERNATIONAL LAW AND ARTIFICIAL INTELLIGENCE

97

formally lawful under certain municipal laws. The phenomenon may also spread rapidly because artificially intelligent entities easily proliferate. One entity can quickly procreate, creating offspring in various jurisdictions. Given these risks, national lawmakers (and scholars) need to examine national law scrupulously, enquiring whether artificially intelligent entities can be created under national law as it currently stands; if yes, whether this possibility should be maintained in the light of possible abuse; if no, how the law should be modified. Restricting the lawfulness of artificially intelligent entities to controlled entities, i.e. to entities which are controlled by human persons who can be held liable, may be part of the answer.21 In addition, States have to review their international legal obligations. One State may go forward and enable the creation of artificially intelligent entities; other States may be bound by treaty law to recognise the legal personality of such entities.22 Within the European Union’s internal market, measures against artificially intelligent entities will be lawful, if the measures specifically target uncontrolled entities with a view to ensuring that a natural person can be held criminally liable when entities have been involved in crimes.23 But outside the Union, in international law, the adoption of such measures may only be lawful after mutual recognition treaties have been changed. If this proves to be the case (and if such measures are desirable), treaty amendment procedures will have to be triggered as soon as possible. Even States that are not under any kind of international obligation to recognise uncontrolled foreign artificially intelligent entities need to decide whether to oppose such foreign entities, since some States will likely tolerate the practice. Irrespective of any legal obligations, the effects of such toleration will be felt worldwide. Flags of convenience and shell companies have clearly demonstrated this. In response, States could try to negotiate a collective response, although this will take time. An internationally agreed verification procedure that would allow a State to learn with sufficient certainty whether a legal entity is under the control of humans should be the priority. As for unilateral measures to be adopted preventively or in case an international response proves unattainable, certain templates do exist, such as the requirement, unilaterally 21

See the distinction between controlled and uncontrolled artificially intelligent entities I propose in Burri (note 17). 22

See LoPucki (note 16), footnote 173.

23

Burri (note 17).

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

98

imposed, to indicate the beneficial ownership of assets in order to forestall money laundering.24 But the effectiveness of such measures would be uncertain. Finally, strict unilateral control mechanisms would risk becoming a hindrance for the myriad of international transactions regular companies and legal entities conduct on a daily basis.

IV. Argument 3: The Geneva Process Will Result in a Ban on Autonomous Weapons Systems, But It Will Be Limited to Weapons Systems Operating Beyond Meaningful Human Control Although international law in general will not be automated (see argument 1), on one level it has been dealing intensively with the possibility of automation. Urged on by non-governmental organisations, a discussion within the forum of the Certain Conventional Weapons Convention25 about a ban on autonomous weapons systems has been ongoing in Geneva since 2013.26 After several preliminary meetings, a group 24 See for details Neil Boister/Robert J. Currie (eds.), Routledge Handbook of Transnational Criminal Law (2015), with chapters on ‘Money Laundering’ by William Gilmore (331–346) and on ‘Transnational Organised Crime’ by Andreas Schloenhardt (409–433). 25 Convention on Prohibitions or Restrictions on the Use of Certain Conventional Weapons Which May Be Deemed to Be Excessively Injurious or to Have Indiscriminate Effects (with Protocols I, II, and III), 10 October 1980, UNTS 1342, 163. 26

A range of issues is explored in Nehal Bhuta et al. (eds.), Autonomous Weapons Systems: Law, Ethics, Policy, (2016); see also Peter Asaro, Jus nascendi, robotic weapons and the Martens Clause, in: Ryan Calo/Michael A. Froomkin/Ian Kerr (eds.), Robot Law (2016), 367; Amnesty International, Autonomous Weapons Systems: Five Key Human Rights Issues for Consideration (2015); Rebecca Crootof, The Killer Robots Are Here: Legal and Policy Implications, Cardozo Law Review 36 (2015), 1837; Jai Galliot, Military Robots: Mapping the Moral Landscape (2015); Paul Scharre/Michael C. Horowitz, An Introduction to Autonomy in Weapon Systems, Center for New American Security Working Paper, February 2015, 23; Geneva Academy of International Humanitarian Law and Human Rights, Autonomous Weapon Systems under International Law, Academy Briefing No. 8, November 2014, 27; Heather Roff, The Strategic Robot Problem: Lethal Autonomous Weapons in War, Journal of Military Ethics 13 (3) (2014), 211; Markus Wagner, The Dehumanization of International Humanitarian Law: Legal, Ethical, and Political Implications of Autonomous Weapon Systems, Vanderbilt Journal of Transnational Law 47 (2014), 1371; Michael N. Schmitt/Jeffrey S. Thurnher, ‘Out of the Loop’: Autonomous Weapon Systems and the Law of Armed Conflict, Harvard National Security Journal 4 (2013), 231; Kenneth Anderson/Matthew C. Waxman, Law and Ethics for Autonomous Weapon Systems: Why a Ban Won’t Work and How the Laws of War Can, 33, available at: https://papers.ssrn. com/sol3/papers.cfm?abstract_id=2250126 (accessed on 27 November 2017); Human Rights Watch/ Harvard International Human Rights Clinic, Losing Humanity: The Case against Killer Robots, November 2012; US Department of Defence, Directive on Autonomy in Weapon Systems, 21 November 2012, available at: http://www.dtic.mil/whs/directives/corres/pdf/300009p.pdf (accessed on 27 No-

INTERNATIONAL LAW AND ARTIFICIAL INTELLIGENCE

99

of governmental experts was instituted and will begin deliberation in November 2017.27 After the notion of the autonomy of weapons systems turned out to be highly controversial among States parties in these discussions, they now seem to converge on ensuring that the use of force in armed conflict, including the use of lethal force, is always subject to meaningful human control. Fully autonomous weapons systems, i.e. systems that select and engage targets without meaningful human control, are likely to be banned by means of a new international legal instrument, while the use of weapons systems equipped with a low level of autonomy will be lawful. A ban on autonomous weapons systems, even if narrowly limited to systems subject to no meaningful human control, would be a significant achievement. It would assuage the fears, which had been widely expressed, that future warfare could be fully automated and that humanity could eventually become subject to the rule of machines.28 However, it would also be insufficient and leave open important points. The term meaningful human control boils down to control. The notion of control, it can be assumed, has garnered support because it is a vague term. It is a longstanding practice in international law to use vague legal terms which all parties can construe according to their preferences when positions diverge widely. Thus, the United States and other States likely to have autonomous weapons systems may interpret control loosely. This allows them to see as lawful a wide variety of constellations, such as an operator who simultaneously controls hundreds of weaponised swarm drones or an armed underwater or air vehicle which has been broadly authorised to use force and loiters for months. Other parties, among whom will be those not in possession of autonomous weapons systems, may take a stricter approach to control. They may expect a tight feedback connection between operator and machine whenever force is used. They thus envisage precisely the opposite of what other States have in mind vember 2017); Ronald C. Arkin, Governing Lethal Behavior in Autonomous Robots (2009); Peter W. Singer, Wired for War: The Robotics Revolution and Conflict in the Twenty-First Century (2009). 27 Fifth Review Conference of the High Contracting Parties to the Convention on Certain Conventional Weapons, Final Document of the Fifth Review Conference, Advance version, 23 December 2016, CCW/CONF.V/10, 9 (decision 1). 28 See Human Rights Watch/Harvard International Human Rights Clinic (note 26); see also European Parliament Committee on Legal Affairs, Draft Report with Recommendations to the Commission on Civil Law Rules on Robotics, 2015/2103(INL), 31 May 2016, introductory clause i: “whereas ultimately there is a possibility that within the space of a few decades AI could surpass human intellectual capacity in a manner which, if not prepared for, could pose a challenge to humanity’s capacity to control its own creation and, consequently, perhaps also to its capacity to be in charge of its own destiny and to ensure the survival of the species”.

100 GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

with regard to ‘control’. In the end, ‘meaningful human control,’ while appearing to be a consensual term, may be nothing more than a charade. Behind the mask of the term the real problems autonomous weapons pose remain unaddressed. Meaningful human control thus fails to make a useful contribution to tackling the issue of symbiosis between man and machine. Arguably, it may be acceptable to turn a blind eye to organic symbiosis in combat. As per now, organic symbiosis seems a rather remote and exotic issue. However, non-organic symbiosis between combatants and machines in concrete situations, such as in inner-city combat where humans collaborate closely with machines, is not remote. It is necessary to know now where the limits of lawful behaviour lie for such situations. Unfortunately, the idea of ensuring meaningful human control does not provide any guidance in this regard. Ensuring that humans retain meaningful control also seems beside the point since the whole concept is at odds with basic premises of warfare. Parties to armed conflicts usually have a strong desire not to give up control over their arms, be it to other conflict parties or machines. This desire sits deep, because giving up control over arms usually equals defeat. Armed forces are, on the contrary, typically designed with a view to retaining control over the use of force and weapons and allocating it, or them, as precisely as possible. (What is more, a party in an armed conflict ready to transfer control over arms to machines is unlikely to take much notice of a ban in any case.) Rather than in ensuring meaningful human control over weapons systems, the principal challenge lies in the use of artificial intelligence and algorithms in warfare. This is where the new arms race takes place.29 The relevant parameters in this race are no longer physical, in contrast to the previous competitions for the most destructive capacity of nuclear weapons or the farthest range of missiles and warplanes. Rather, the parameters relate to data, as in the resolution of satellite images and the accuracy of digital maps. This dimension, its implications for the future balance of power, and its application in online warfare require closer attention. At worst, establishing regulation of meaningful human control over weapons systems distracts from them. This may prove disastrous, since deep and thick connections to civilian applications,

29 See Tom Simonite, For Superpowers, Artificial Intelligence Fuels New Global Arms Race, Wired, 8 September 2017, available at: https://www.wired.com/story/for-superpowers-artificial-intelligencefuels-new-global-arms-race/ (accessed on 27 November 2017).

INTERNATIONAL LAW AND ARTIFICIAL INTELLIGENCE

101

such as autonomously driven vehicles or image recognition, and scalability make the topic tough to regulate in the best of circumstances.

V. Argument 4: Existing International Law Offers Valuable Insights into the Meaning of Control Over Artificial Intelligence and the Limits of Delegation From argument 2 and argument 3 a central point emerges, namely that control over artificial intelligence is a key dimension. Artificially intelligent entities may be considered to operate lawfully as long as they are subject to human control; autonomous weapons systems may soon become unlawful, unless they are controlled by humans. Whereas the law directly concerned by these manifestations of artificial intelligence (company law and the law of armed conflict, respectively) includes indications of how control issues can be solved – think of the laws regulating a company board’s control over management or the chain of command in armed forces – international law has available a rich jurisprudence with respect to control, which has accumulated over the years in the most diverse situations. This experience should be drawn upon to shed some light on the puzzling control issues associated with artificial intelligence. The most obvious source in this regard is the case law of the international criminal courts. Many of the cases before the International Criminal Tribunal for the Former Yugoslavia deal precisely with control: When and under which conditions is a commander high up in the hierarchy responsible for deeds on the ground? When are instructions sufficiently precise to warrant attribution? When are the tasks divided among several actors so strongly linked that they may be considered as one entity, a joint criminal enterprise?30 The answers this case law gives to such questions may contain clues of how control over artificial intelligence can be practically tackled. The fact that in international criminal law humans are exclusively involved, while with artificial intelligence one or more humans interact with a machine, should not be a reason not to draw on the tribunal’s record. Communication theory, at least, indi-

30

Answers to these questions can be found in many decisions of the ICTY. By way of example, see ICTY, Trial Chamber, The Prosecutor v. Šainovic et al., IT-05-87, Judgment of 26 February 2009, and id., Appeals Chamber, Šainovic et al. (note 30), Judgment of 23 January 2014.

102 GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

cates that correspondence with an artificially intelligent agent need not necessarily be fundamentally different from social intercourse among humans.31 General international law may not speak to control as directly as international criminal law. However, the situations the World Court has addressed are even richer, more diverse, and thus more informative.32 Consider the following two examples stemming from the time of the League of Nations. They are just a few among dozens. i) In 1931, Austria had concluded a treaty with Germany establishing a customs union. Soon thereafter, the Permanent Court of International Justice was asked to give an opinion on whether Austria violated its international obligation laid down in previous treaties (inter alia the peace treaty of Saint-Germain-en-Laye of 10 September 1919) not to alienate its independence. The Court concluded that Austria had partly violated this obligation.33 The opinion provides an illustration of what it means not to alienate one’s independence. Is this not also relevant for humans who are prone to alienate their independence and subject themselves to the ‘will’ of artificial intelligence? ii) In 1930, the same Court indicated in an Opinion that the Free City of Danzig was precluded by its status that was secured inter alia in its constitution, which in turn was guaranteed by the League of Nations, from joining the International Labour Organization.34 The following quotes from the opinion are evidence of the Opinion’s relevancy: “The result is that, as regards the foreign relations of the 31

Gunther Teubner’s talk given in Zurich on 11 September 2017 on legal personhood indicated this, Gunther Teubner, Digitale Rechtssubjekte? Zum privatrechtlichen Status autonomer informationstechnischer Systeme, presentation held at the Tagung der Zivilrechtslehrervereinigung, Zurich, 10– 12 September 2017. 32 International law scholarship also addressed control and delegation in various circumstances (even beyond autonomous weapons systems – for the latter see supra, note 26). See for instance Erik Voeten, Delegation and the Nature of Security Council Authority, in: Bruce Cronin/Ian Hurd (eds.), The UN Security Council and the Politics of International Authority (2008), 43. Elements of control and delegation are also present in the discussions about constitutionalism and pluralism. By way of example see Bardo Fassbender, The United Nations Charter as Constitution of the International Community, Columbia Journal of Transnational Law 36 (3) (1998), 529, and most recently id., International Constitutional Law: Written or Unwritten?, Chinese Journal of International Law 15 (3) (2016), 489; Nico Krisch, Beyond Constitutionalism: the Pluralist Structure of Postnational Law (2012); Anne Peters, Compensatory Constitutionalism: The Function and Potential of Fundamental International Norms and Structures, Leiden Journal of International Law 19 (2006), 579. 33 Permanent Court of Justice (PCIJ), Customs Régime Between Germany and Austria, Advisory Opinion of 5 September 1931, Series A/B, No. 41. 34

Id., Free City of Danzig and International Labour Organization, Advisory Opinion of 26 August 1930, Series B, No. 18.

INTERNATIONAL LAW AND ARTIFICIAL INTELLIGENCE

103

Free City, neither Poland nor the Free City are completely masters of the situation.”35 And: “[…S]o far as these rights involve a limitation on the independence of the Free City, they constitute organic limitations which are an essential feature of its political structure.”36 Some questions follow naturally, namely what does it mean to be the complete master of a situation involving artificial intelligence or what are the organic limitations of our artificial intelligence’s structure? Overall, the opinion offers an illustration of how to structure a situation of competing interests and mutual dependency – which is just the point to be addressed with regard to artificial intelligence. The flipside of retaining control over something is that some decisions must not be delegated. The persons controlling an artificial intelligence necessarily need to take some decisions themselves, else they would not be in control. The questions to be answered then are: Which decisions are these? What is it that cannot be delegated to a machine? With regard to autonomous weapons systems, the consensus seems to be that the decision to kill a human person in a concrete combat situation cannot be delegated to a machine. With regard to the artificially intelligent entities discussed in argument 2, the answer may be that a human person needs to be chargeable in case crimes are committed; criminal responsibility thus cannot be delegated.37 A search for other such limits of delegation in international law reveals again an interesting decision by the Permanent Court of International Justice, namely Consistency of Certain Danzig Legislative Decrees With the Constitution of the Free City.38 The legislative organs of Danzig had introduced a very general penal norm, authorising authorities to sanction individuals when an act deserved a penalty according to fundamental conceptions of penal law and sound popular feeling. The Court in 1935 advised that such a norm, in moving beyond the nullum crimen principle, violated the fundamental rights of individuals and the rule of law. In the perspective of artificial intelligence, the ruling suggests that the discretion delegable to an individual – which in the case of the Permanent Court was an individual judge, while in the case of artificial intelligence it is a synthetic individual – may be limited. The orders the human principal gives may have to be specific, the space the digital agent can fill may have to be lim35

Ibid., 13.

36

Ibid., 11.

37

See Burri (note 17).

38

PCIJ, Consistency of Certain Danzig Legislative Decrees With the Constitution of the Free City, Judgment of 4 December 1935, Series A/B, No. 65.

104 GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

ited. This leads to the question: Should we commit to the basic principle that artificial intelligence must be precluded from taking discretionary decisions?39 Certain Danzig Legislative Decrees and the other cases discussed above, thus, hint at the broader task ahead for lawyers. Like the foregoing international case law, the human rights case law of the European Court of Human Rights, which is much thicker, is likely full of the kinds of clues offered by Certain Danzig Legislative Decrees. In addition to the overarching perspective on delegation and control in general, guidance can be drawn from the case law with regard to each specific fundamental right. In the same way that Certain Danzig Legislative Decrees speaks to nullum crimen, thus enabling inference for artificial intelligence, the case law of the European Court of Human Rights speaks to the right to life, the prohibition of torture, etc., and in each case implications for artificial intelligence are likely. Scrutinising the case law of the European Court of Human Rights is a mammoth task (not to think of the case law of national courts!). However, in addition to exposing the limits of delegation and control, it will become evident where our legal systems are vulnerable to the artificial intelligence revolution under way. Data protection is already further ahead in coming to grips with the consequences of artificial intelligence because it threatens digital privacy most directly. But the law governing companies, contracts, banking, agency – not to speak of public law in general – is nowhere near that far. It therefore seems urgent to start to look at the case law through the lens of artificial intelligence.

39

Like the three decisions of the PCIJ discussed in this section the following decisions of this court may prompt further questions with regard to artificial intelligence: PCIJ, Jaworzina (Polish-Czechoslovakian Frontier), Advisory Opinion of 6 December 1923, Series B, No. 8, (does ejus est interpretare legem cujus condere imply anything for artificial intelligence? (at 37)); re powers and delegation of decisions: id., The Monastery of Saint-Naoum (Albanian Frontier), Advisory Opinion of 4 September 1924, Series B, No. 9; id., Jurisdiction of the European Commission of the Danube Between Galatz and Braila, Advisory Opinion of 8 December 1927, Series B, No. 14, and id., Interpretation of the Greco-Turkish Agreement of December 1st, 1926, Advisory Opinion of 28 August 1928, Series B, No. 16; id., Rights of Minorities in Upper Silesia (Minority Schools), Judgment of 26 April 1928, Series A, No. 15, and id., Access to German Minority Schools in Upper Silesia, Advisory Opinion of 15 May 1931, Series A/B, No. 40 (what does the fact that certain decisions must not be delegated or questioned, e.g. the decision to belong to a minority, imply for artificial intelligence?); id., Interpretation of the Statute of the Memel Territory, Judgment of 11 August 1932, Series A/B, No. 49 (what does it mean to have autonomy?); id., Legal Status of Eastern Greenland, Judgment of 5 April 1933, Series A/B, No. 53 (how does one control something?) – and these are just the decisions of the PCIJ.

INTERNATIONAL LAW AND ARTIFICIAL INTELLIGENCE

105

VI. Argument 5: Supersoft Law Will Govern Artificial Intelligence Companies seeking profit, sometimes joined by States in quest of security and power, drive forward the technology of artificial intelligence. Autonomous vehicles, image recognition, advisory algorithms, financial technology, and autonomous weapons systems are all evidence of this drive. However, wealthy individuals and scientists have cautioned against the risks involved in developing artificial intelligence,40 and public institutions41 have echoed the call. Consequently, a number of initiatives have been launched which aim to explore the ethical and societal implications of artificial intelligence.42 Researchers and groups of citizens are poring over a diverse set of issues, e.g. the ethical considerations to be taken into account by arms procurement agencies,43 the implications of artificial intelligence personhood,44 new digital fundamental rights,45 etc. The Institute of Electrical and Electronics Engineers (IEEE), a technical professional society, for instance, pushes hard to standardise ethical principles. Its Global Initiative for Ethical Considerations in Artificial Intelligence and Autonomous Systems recently published the first version of a charter entitled ‘Ethically Aligned Design’46 in the creation of which more than 100 experts were involved. It has since 40 Future of Life Institute, An Open Letter: Research Priorities for Robust and Beneficial Artificial Intelligence, available at: http://futureoflife.org/ai-open-letter/ (accessed on 27 November 2017). 41

European Parliament Committee on Legal Affairs (note 28).

42

E.g. the Future of Life Institute, available via: www.futureoflife.org (accessed on 27 November 2017) and the AI Initiative at the Future Society, available via: http://ai-initiative.org (accessed on 27 November 2017). See the so-called Asilomar AI Principles, available at: https://futureoflife.org/aiprinciples/ (accessed on 27 November 2017). N.b. how the Principles were created: Future of Life Institute, A Principled AI Discussion in Asilomar, 17 January 2017, available at: https://futureoflife. org/2017/01/17/principled-ai-discussion-asilomar/ (accessed on 27 November 2017). 43

E.g. a research group at the University of Zurich, led by Markus Christen, is currently developing an assessment scheme for the Swiss arms procurement agency. 44

Bendert Zevenbergen at University of Princeton in April 2017 convened a workshop on artificial intelligence personhood. See also Bayern et al. (note 15). 45

See the proposed charter of digital fundamental rights, available via: www.digitalcharta.eu (accessed on 27 November 2017). 46

Institute of Electrical and Electronics Engineers (IEEE), Ethically Aligned Design, The IEEE Global Initiative for Ethical Considerations in Artificial Intelligence and Autonomous Systems, 13 December 2016, available at: http://standards.ieee.org/develop/indconn/ec/ead_v1.pdf (accessed on 27 November 2017).

106 GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

established ten public working groups on standards.47 They address a whole range of issues, e.g. transparency, algorithmic bias, and privacy. The aim is to create general ethical standards to be applied by designers of systems, including programmers of artificial intelligence, makers of robots, etc., in all situations of human-machine interaction, including autonomous driving, care giving, financial investment, armed combat, etc. The IEEE’s universal standards aim to be applied around the world, from China to the United States to Europe. The numerous processes currently under way are creating international standards. But this happens outside traditional international law-making fora, which are bypassed. States, which are the traditional subjects and makers of international law, are only involved in the Geneva process addressing autonomous weapons systems (see argument 3). The United Nations has, at least so far, only played a marginal role. Instead, an amorphous and leaderless legislature is taking on standard making. It involves interested individuals, professional associations, social and natural scientists, companies, and civil society organisations. Individuals gather at universities; debates are taking place online. The standards that are being produced also defy the categories of international law. They will not formally be binding, but they will not amount to soft international law, either, since none of the traditional subjects or bodies of international law will have been involved in their making. The standards will be a kind of supersoft law: Emerging out of the international legal vacuum, they will be persuasive on their merits and imbued with a strong compliance pull, despite their non-binding form. 47 The IEEE’s recently established working groups on standards are: P7000, Model Process for Addressing Ethical Concerns During System Design, available at: http://sites.ieee.org/sagroups-7000/; P7001, Transparency of Autonomous Systems, available at: https://standards.ieee.org/develop/project/ 7001.html; P7002, Data Privacy Process, available at: https://standards.ieee.org/develop/project/7002. html; P7003, Algorithmic Bias Considerations, available at: https://standards.ieee.org/develop/project/ 7003.html; P7004, Standard on Child and Student Data Governance, available at: https://standards. ieee.org/develop/project/7004.html; P7005, Standard for Transparent Employer Data Governance, available at: https://standards.ieee.org/develop/project/7005.html; P7006, Standard on Personal Data AI Agent, available at: https://standards.ieee.org/develop/project/7006.html; P7007, Ontological Standard for Ethically Driven Robotics and Automation Systems, available at: https://standards.ieee.org/ develop/project/7007.html; P7008, Standard for Ethically Driven Nudging for Robotic, Intelligent and Autonomous Systems, available at: http://standards.ieee.org/develop/project/7008.html; P7009, Standard for Fail-Safe Design of Autonomous and Semi-Autonomous Systems, available at: http:// standards.ieee.org/develop/project/7009.html; and P7010, Wellbeing Metrics Standard for Ethical Artificial Intelligence and Autonomous Systems, available at: http://standards.ieee.org/develop/project/ 7010.html (all accessed on 27 November 2017).

INTERNATIONAL LAW AND ARTIFICIAL INTELLIGENCE

107

Not all of this is new. Janet Koven Levit described three processes of “bottom-up law-making” and dubbed it a myth that States today made international law.48 The internet was also created through a process of standardisation driven by forces beyond the horizons of traditional international law. The Internet Corporation for Assigned Names and Numbers has never properly fitted into the international legal landscape. Its transformation in October 2016 goes beyond the scope of States and international organisations.49 Code, more generally, is a universal language that standardises all computer operations.50 What seems new, though, is that the standardisation reaches beyond the technical. The ethical, the moral, and the political will no longer be subliminal and dealt with incidentally as before in standards but will take the centre stage. Standards will address moral and ethical questions, which had hitherto been the domain of thinkers and philosophers, and political questions, which had previously been settled through political discourse. What seems to be going on currently is nothing less than the universal standardisation of morals and ethics.51

VII. Conclusion International law will not be automated, unlike some domains of domestic law. This, however, does not mean that international lawyers can afford to ignore artificial intelligence. Its impact will be felt widely. Artificially intelligent entities – i.e. legal persons established under municipal law that ‘house’ artificial intelligences – will, for instance, emerge and prompt international law to adapt. Autonomous weapons 48

Janet Koven Levit, Bottom-Up International Lawmaking: Reflections on the New Haven School of International Law, The Yale Journal of International Law 32 (2007), 393. See also the case studies in Andreas Fischer-Lescano/Gunther Teubner, Regime-Collisions: The Vain Search for Legal Unity in the Fragmentation of Global Law, Michigan Journal of International Law 25 (2003–2004), 999, 1017 et seq. 49

See Anon., We the networks, The Economist, 5 March 2016, available at: https://www.economist. com/news/international/21693922-organisation-runs-internet-address-book-about-declare-independencewe (accessed on 27 November 2017). 50 51

See Lawrence Lessig, Code and Other Laws of Cyberspace (1999).

The new standardisation, however, does not encompass everything. It does not imply that the traditional ways in which international law operates become useless. Standardisation does not contribute to the solution of the South China Sea dispute(s) any more than it solves world trade disputes, plays a part in the prosecution of international criminals, or shows a way out of the conundrums involving Kosovo, Catalonia, etc. These processes do not lend themselves to standardisation any more than to the automation discussed in argument 1.

108 GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

systems currently enjoy international lawyers’ undivided attention, but at its core the challenge autonomous weapons raise is broader. It is one of general control and delegation. This becomes evident in artificially intelligent entities, too. International law has a long history of dealing with control and delegation. The case law of international courts abounds with issues of control and delegation. By way of example, this article looked at three decisions of the Permanent Court of International Justice, but the case law of the International Court of Justice, the international criminal courts, and the regional human rights courts is likely to teem with these issues. International law thus offers inspiration and illustration for how artificial intelligence can and should be governed. Yet governance of and by artificial intelligence is at the present time in the hands of informal standardisers, such as the IEEE. International lawyers should step up their game.

Testing International Legal Regimes: The Advent of Automated Commercial Vessels ALDO CHIRCOP(

ABSTRACT: International shipping is on the eve of a new era where remotely controlled and partially or fully automated and unmanned Maritime Autonomous Surface Ships (MASS) will be carrying international trade. The regulation of navigation and shipping in the contemporary international law of the sea and international maritime law are premised on human presence and control on-board ships. Provisions of the United Nations Convention on the Law of the Sea of 1982 and several maritime conventions will need to be revisited to determine how MASS may be accommodated and, where not possible, what further legal development may be needed. Recently, the International Maritime Organization (IMO) decided to address the expected regulatory impacts of these ships and to prepare an agenda for their proactive regulation. This article explores regulatory impacts that would need to be considered and argues that MASS have the potential to provide new directions for international law and the IMO. KEYWORDS: International Law of the Sea, International Maritime Law, International Maritime Organization, Maritime Autonomous Surface Ships, Unmanned Ships

I. Introduction In recent years, research in artificial intelligence and ship technology has advanced to such an extent as to pave the way for partial or full autonomous commercial shipping. Technology has always been at the centre of shipping regulation and is frequently a major driver for new standards. Recent technological and commercial triggers are leading major industry actors to explore different degrees of remotely-controlled ship voyages1 and autonomous ship operations for the near future.2 Some port operations ( Professor of Law and Canada Research Chair in Maritime Law and Policy, Marine & Environmental Law Institute, Schulich School of Law, Dalhousie University, Canada. 1 Rolls-Royce and towage company Svitzer claim to have performed the first remotely-controlled vessel operation. See Anon., World’s First Remotely-Controlled Commercial Vessel Put to the Test in

110

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

appear to be also planning to embrace automation.3 The absence of a crew on a vessel, per se, is not a new development, as unmanned towed barges and remotely operated submersibles for research and offshore petroleum development have long been in use. What is new is a combination of passive presence or even absence of on-board crew, remotely-controlled ships, and partial or full autonomous decision-making through instrumentation. These technologies are anticipated not only for new builds, but also for vessels retrofitted for remote control operations.4 The range of ships affected is diverse.5 Copenhagen, gCaptain, 20 June 2017, available at: http://gcaptain.com/worlds-first-remotelycontrolled-commercial-vessel-put-to-the-test-in-copenhagen/?utm_source=feedburner&utm_ medium=feed&utm_campaign=Feed%3A+Gcaptain+%28gCaptain.com%29 (accessed on 18 September 2017). Wei Zhe Tan, NYK aims to pilot remote-controlled boxship in 2019, Lloyd’s List, 24 August 2017, available at: https://lloydslist.maritimeintelligence.informa.com/LL111056/NYK-aims-to-pilotremotecontrolled-boxship-in-2019 (accessed on 14 December 2017). See also Michelle Howard, Wärtsilä Tests Remote Control Ship Operating Capability, Marine Link, 18 September 2017, available at: https://www.marinelink.com/news/capability-operating428998?utm_source=MT-ENews-2017-0901&utm_medium=email&utm_campaign=MT-ENews (accessed on 18 September 2017). 2

The Unmanned Cargo Ship Development Alliance plans to deliver the first unmanned cargo ship in 2021. The Alliance is led by Chinese HNA Technology Group Co and includes CCS, China Ship Research & Development Institute, Shanghai Marine Diesel Engine Research Institute, HudongZhonghua Shipbuilding (Group) Co., Marine Design Research Institute of China (MARIC), RollsRoyce, Wärtsilä, and the American Bureau of Shipping (ABS), Anon., ABS Joins Alliance Developing Unmanned Cargo Ship, gCaptain, 26 July 2017, available at: http://gcaptain.com/abs-joins-alliance-todevelop-unmanned-cargo-ship/ (accessed on 14 December 2017). A 120 Twenty-foot Equivalent Units autonomous open top fully-electric and zero emission container ship is under construction, see Kongsberg Maritime, Autonomous ship project, key facts about YARA Birkeland, available at: https://www. km.kongsberg.com/ks/web/nokbg0240.nsf/AllWeb/4B8113B707A50A4FC125811D00407045? OpenDocument (accessed on 18 September 2017). Bourbon, Automated Ships Ltd, and Kongsberg Gruppen ASA plan to build autonomous offshore service vessels, see Anon., Bourbon Joins Project to Build World’s First Autonomous Offshore Vessel, gCaptain, 11 July 2017, available at: http://gcaptain. com/bourbon-joins-project-to-build-first-autonomous-offshore-vessel/ (accessed on 18 September 2017). 3 Anon., VCIT world’s first fully automated container terminal: Cargotec, Canadian Shipper, 11 September 2017, available at: https://www.canadianshipper.com/transportation-and-logistics/vcit-worldsfirst-fully-automated-container-terminal-cargotec/1003374728/ (accessed on 18 September 2017). 4

Alan M. Weigel/Sean T. Pribyl, The Future Is Now: Unmanned and Autonomous Surface Vessels and Their Impact on the Maritime Industry, Mainbrace, June 2017, 22, 23, available at: http://www. blankrome.com/siteFiles/Publications/Mainbrace_June_2017.pdf (accessed on 18 September 2017). 5

In particular, the carriage of bulk cargoes is expected to be a prime candidate for autonomous ships, see David Stringer, Autonomous Cargo Ships Extend Miner’s Technology Drive to Seas, gCaptain, 7 June 2017, available at: http://gcaptain.com/autonomous-cargo-ships-extend-miners-technology-driveseas/?utm_source=feedburner&utm_medium=feed&utm_campaign=Feed%3A+Gcaptain+%28 gCaptain.com%29 (accessed on 18 September 2017). However, automation on containerships is also anticipated. NYK is reported to be planning to test remote controlled steering of a boxship with a stand-by crew, Anon., NYK to Test Autonomous Box Ship in the Pacific Oceans in 2019, World Maritime News,

AUTOMATED COMMERCIAL VESSELS

111

The history of international maritime law is punctuated by game-changing technological milestones. This is not surprising as the ship is a platform for varied instrumentation, with the consequence that technological development serves to enhance ship function and operations and, in turn, rules and standards for safety, security, and environment protection. The advent of the modern cruise liner and early ship losses, most especially the Titanic, led to the adoption of the first convention on safety of life at sea.6 In the mid-19th century, the first rules of the road were introduced as a result of the introduction of the steam engine in shipping and the need to reduce collisions at sea. Following World War II, the widespread introduction of radar on commercial vessels led to numerous radar-assisted collisions which resulted in important changes in the application of the rules of the road.7 During the same period, on-board freezer technology enabled the transportation of refrigerated cargoes and in the process produced new trades.8 Within a short period thereafter, the advent of the supertanker and containerisation revolutionised the carriage of goods in bulk and packaged forms. The increased use of tankers led to far-reaching vessel-source pollution regulation in both the public and private law spheres. This extended to construction and operational requirements9 as well as liability regimes.10 Container vessels transformed maritime trade and multi-modal carriage of cargo.11 More recently, with the receding of summer sea ice as a result of climate change and growth of shipping in polar regions, 25 August 2017, available at: http://worldmaritimenews.com/archives/228202/nyk-to-test-autonomousboxship-in-2019/ (accessed on 18 September 2017). 6 Eventually leading to the current iteration, the International Convention for the Safety of Life at Sea, 1 November 1974, UNTS 1184, 2 (SOLAS Convention). 7 In particular the rules on look out and safe speed. Arts. 5–6 Convention on the International Regulations for Preventing Collisions at Sea, 20 October 1972, UNTS 1050, 16 (COLREGS). 8

See Stanley H. Beaver, Ships and Shipping: The Geographical Consequences of Technological Progress, Geography 52 (1967), 133. 9

Primarily under the International Convention for the Prevention of Pollution from Ships, 2 November 1973, UNTS 1340, 184, as amended by the Protocol Relating to the International Convention for the Prevention of Pollution from Ships of 1973, 17 February 1978, UNTS 1340, 61 (MARPOL). Annex 1 includes standards and rules for construction and discharge of oily wastes from tank washing at sea. 10 In particular the following: International Convention on Civil Liability for Oil Pollution Damage, 29 November 1969, UNTS 973, 3 (CLC Convention); International Convention on the Establishment of an International Fund for Compensation for Oil Pollution Damage, 18 December 1971, UNTS 1110, 57 (IOPCF Convention). 11

See Marc Levinson, The Box: How the Shipping Container Made the World Smaller and the World Economy Bigger (2006).

112

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

new rules for the construction, equipping, crewing, and operations for vessels navigating those waters have been adopted.12 The technology of shipping has driven much of the opus of the International Maritime Organization (IMO), a specialised agency of the United Nations and the competent organisation with respect to international shipping.13 Maritime autonomous surface ships (MASS) have the potential of triggering new regulatory responses and the IMO is yet again expected to lead that exercise. Across the field of maritime regulation, a source of constant concern is the human factor in vessel operations. Autonomous shipping has the potential to substantially reduce or even remove the human factor in the operation and navigation of the ship, relegating it to a passive or temporary on-board presence and even to a fully shorebased monitoring and controlling role. The role of automation ranges from supporting human decision-making to autonomous decision-making. In setting standards and rules for shipping, maritime regulation has frequently been concerned with the influence of human decision-making and conduct, by employing technologies and processes to produce safety, environmental, and security outcomes. Autonomous shipping has the potential to transfer human assessment and decision-making to artificial intelligence. This entails the use of “algorithms, tools and techniques that mimic human learning to solve specific problems” and “to recognize patterns in data, making predictions from previously unseen data.”14 These systems will be able to gen12

International Association of Classification Societies (IACS), Requirements Concerning Polar Class (2011), available at: http://www.iacs.org.uk/document/public/Publications/Unified_require ments/PDF/UR_I_pdf410.pdf (accessed on 18 September 2017). The IACS polar class requirements are implicitly adopted in the Polar Code, Part I-A para. 1.2.10. The Code and related amendments to international conventions were adopted in stages: Maritime Safety Committee (MSC), International Code for Ships Operating in Polar Waters (Polar Code), MSC Resolution 385(94) of 21 November 2014, Report on its 94th Session, IMO Doc. 94/21/Add.1 (2014), Annex 6; id., Amendments to the International Convention for the Safety of Life at Sea of 1974, as amended, in ibid., Annex 7. For a consolidated online edition of the Polar Code, see International Code for Ships Operating in Polar Waters, available at: http://www.imo.org/en/MediaCentre/HotTopics/polar/Documents/POLAR% 20CODE%20TEXT%20AS%20ADOPTED.pdf (accessed on 18 September 2017). 13 The United Nations Convention on the Law of the Seas (UNCLOS, see infra, note 26) designates the International Maritime Organization (IMO) as the competent international organisation with respect to international shipping in numerous provisions. These are listed in “Competent or Relevant International Organizations” under the United Nations Convention on the Law of the Sea, Law of the Sea Bulletin 31 (1996), 79, 81 et seq. 14 “Machine Learning is a set of algorithms, tools, and techniques that mimic human learning to solve specific problems. By analyzing existing data sets, machine learning can be used to recognize patterns in data, making predictions from previously unseen data. The bigger the data set, the more complex the patterns the model can recognize and the more accurate the predictions.”, Anon., Rolls-Royce

AUTOMATED COMMERCIAL VESSELS

113

erate and compute more data and significantly enhance situational awareness, but will they be a replacement for human judgement? There are good reasons to believe that unmanned and autonomous ships will not be technologies of passing interest. There is a discernible declining interest in seafaring careers with a consequent trend towards shortage of seafarers. As ships have grown bigger, crews have become smaller, working long hours and frequently performing tedious tasks, and having little if any time for port visits because of fast cargo turnaround and security reasons. Crewing costs are one of the most significant elements in a ship’s cost structure.15 Without a crew (at least permanent) and in the absence of a need of full crew quarters and a design to enable safe movement on board, the autonomous ship will be lighter and cheaper to build. Most passageways for crews will be removed. Crew quarters will be minimised and perhaps even removed. When needed, crew space may be containerised and installed temporarily into the ship. The bridge as conceived in contemporary ships will be smaller and simpler. In addition to saving on crew salaries, there will be savings on crew supplies and energy. The ship will have a slick design and will be equipped with a complex array of sensors and instrumentation. Its design and the absence of a crew will make it difficult for pirates to board. It will likely be navigated more accurately and consequently with fewer errors, if any, than crewed vessels. It will consume less energy and produce fewer greenhouse gas emissions. However, the deployment of autonomous ships could produce new concerns, in particular with respect to the interaction with crewed ships in their vicinity and cyber security.16 The IMO’s Maritime Safety Committee (MSC) has heeded a recent call to launch a regulatory scoping exercise with respect to ‘maritime autonomous surface ships’ to Use Google Machine Learning in Quest for Autonomous Ships, gCaptain, 3 October 2017, available at: http://gcaptain.com/rolls-royce-to-use-google-machine-learning-in-quest-for-autonomous-ships/ (accessed on 3 October 2017). 15 In 2015, crew costs were the third highest cost at 17% after ship finance and competition costs, each at 22%. See Anon., Ship Operating Costs Set to Rise, Maritime Executive, 30 October 2015, available at: http://maritime-executive.com/article/ship-operating-costs-set-to-rise (accessed on 18 September 2017). 16 The concern is that autonomous ships may become targets for hackers. In the recent hacking of MAERSK, the company is reported to have lost $200–300 million. Danish insurer Tryg predicts that 90% of its clients will buy cyber insurance, Stine Jacobsen, Rising Hacker Threat Will Trigger Boom in Cyber Crime Insurance, Tryg Says, Reuters, 4 October 2017, available at: https://www.reuters.com/ article/us-tryg-cyber/rising-hacker-threat-will-trigger-boom-in-cyber-crime-insurance-tryg-saysidUSKCN1C91MV (accessed on 4 October 2017).

114

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

(MASS), the term used in this article to capture the emerging technologies of unmanned and autonomous ships.17 A multi-Member State submission highlighted the undesirability of leaving the regulation of autonomous ships to the national level in the absence of an appropriate international framework.18 After all, as ‘the competent international organisation’ with respect to international shipping matters, it is the responsibility of the IMO to develop and maintain the international standards and rules necessary for safe navigation. The IMO has been requested to identify regulations which preclude unmanned operations, regulations that do not apply to unmanned operations because of the absence of a crew, and regulations which do not preclude unmanned operations but which may need to be amended to take into consideration the operation of MASS.19 Against this backdrop, it is appropriate to enquire whether the advent of MASS requires mere incremental or adaptive regulation, or rather constitutes a gamechanger requiring substantial new regulatory directions, and further what might be consequences for the international law of the sea with respect to provisions concerning navigation. This inquiry explores what could or should be expected regulatory impacts when the human factor is reduced or eliminated from decision-making with respect to ship operations where traditionally human presence and control are presumed, required, and regulated with attendant legal consequences. There is an emerging body of scholarship on the public and private law aspects of the subject and it raises important questions with respect to the potential impacts in both fields.20 The 17 MSC, Report of the Maritime Safety Committee on its 98th Session, IMO Doc. MSC 98/23 (2017), 79. 18 Maritime Autonomous Surface Ships Proposal for a Regulatory Scoping Exercise, Submitted by Denmark, Estonia, Finland, Japan, the Netherlands, Norway, the Republic of Korea, the United Kingdom and the United States, IMO Doc. MSC 98/20/2 (2017) (MASS Scoping Exercise). See also Impact of New and Advancing Technologies to Maritime Transport and the Regulatory Framework, Submitted by Denmark, Estonia, Finland, Japan, Norway, Singapore, Sweden and IMarEST, IMO Doc. MSC 98/22/7 (2017). 19 20

Ibid.

For example: Eric Van Hooydonk, The Law of Unmanned Merchant Shipping: An Exploration, Journal of International Maritime Law 20 (2014), 403; Paul W. Pritchett, Ghost Ships: Why the Law Should Embrace Unmanned Vessel Technology, Tulane Maritime Law Journal 40 (2015–2016), 197; Hannah Stones, Objective and Subjective Safety in Unmanned Shipping, Shipping and Trade Law (STL) 16 (9) (2016), 4; Trudi Hogg/Samrat Ghosh, Autonomous Merchant Vessels: Examination of Factors that Impact the Effective Implementation of Unmanned Ships, Australian Journal of Maritime and Ocean Affairs 8 (2016), 206; Michal Chwedczuk, Analysis of the Legal Status of Unmanned Commercial Vessels in U.S. Admiralty and Maritime Law, International Journal of Machine Learning

AUTOMATED COMMERCIAL VESSELS

115

focus of this article is on public law aspects. This author argues that MASS have the potential to provide new directions for international law and the IMO. As the MSC embarks on its scoping exercise during the expected four sessions, this article explores the extent to which MASS are or can be accommodated in contemporary regulation and explores areas of potential direct and indirect impacts. The article concludes with reflections on how well-positioned maritime regulation is, or is not, with respect to the withdrawal of the human factor from shipping in favour of automated systems and the possible implications for further legal development.

II. Technology and Terminology While the term used in this article is MASS, the terminology of autonomous ships is not yet settled. The term ‘MASS’ appears in the initial discourse at the IMO, whereas industry circles and scholarly literature have used other diverse terms, including ‘unmanned surface vessels’, ‘autonomous unmanned merchant vessels,’ ‘robot ships’, ‘crewless ships’, ‘unmanned ships’, and ‘automated ships’, among others. In considering how MASS will impact contemporary maritime regulation, an early task for the IMO will be terminological clarity and consistency. Terminology has already triggered varying views, with some Member States preferring to define the term to clarify the scope of the Organization’s future work, whereas others preferred to leave definitional matters to a later stage.21 Automation in shipping is not a complete novelty. Various aspects of a ship’s operations have been subject to automation for some time, such as automatic coursesetting and certain machinery functions in the engine room. MASS will navigate partially or fully without a master and crew on board and with its control emanating from an automated system or through remote means. The ship will be equipped with instrumentation and software programmes that will ensure its accurate location, using global positioning (GPS) and automatic identification system (AIS) transmission, and perform various functions. It will have situation awareness through sensors and Cybernetics 47 (2016), 123; Robert Veal/Michael Tsimplis, The Integration of Unmanned Ships into the Lex Maritima, Lloyd’s Maritime and Commercial Law Quarterly (2017), 303; Robert Veal, Unmanned Ships on the IMO Work Agenda, STL 17 (5) (2017), 1. 21

MSC (note 17), 79.

116

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

and cameras that will serve as its eyes and ears, sensing the presence of other vessels, as well as the usual radar so as to navigate in compliance with the rules of the road.22 The vessel will be generating and receiving digital data via satellite communication. Algorithms will be at the heart of computerised autonomous decision-making for safe navigation. This technology may be employed to assist the crew on board or even to replace them, whether temporarily or for much of, if not the whole voyage. The term ‘unmanned’ does not necessarily entail the total absence of human intervention in the navigation and operation of the vessel at sea. A concept advanced by a major research project considers how the technology may be introduced with respect to bulk cargo carriage and with safety outcomes at least comparable to manned ships.23 As bulkers have long voyages usually without multiple port cargo drops, they would need a temporary crew and pilot only during the port approach, berthing, and de-berthing. The long unmanned voyage would be monitored remotely from a shorebased control centre. Thus the ship may operate autonomously for periods of time, but humans will be on standby for ready intervention if or when needed, or the vessel may be remotely controlled and operated when needed. Unlike the autonomous vessel, the remotely controlled vessel is directed from a shore base. Thus, the traditional fully manned vessel and the concept of the fully autonomous ship are two ends of a spectrum along which there is a range of vessels with differing degrees of manning, remote control, and automation. A Danish study from an engineering perspective describes the degrees of ship automation beyond manual navigation to include, in order of increasing intensity, automatic course steering (on autopilot on the basis of a human encoded course), decision support (the system provides information for a human decision), remotely operated navigation (off-board instructions for autopilot and machinery), remote monitoring (full off-board situational ship awareness – radar, digital, TV-monitoring – to enable remote controlled navigation), partial autonomy (on-board partial decision-making autonomy to inform on-

22

Thomas Porathe/Johannes Prison/Yemao Man, Situation Awareness in Remote Control Centres for Unmanned Ships, paper presented at Human Factors in Ship Design & Operation, London, 26–27 February 2014, available at: http://publications.lib.chalmers.se/records/fulltext/194797/local_ 194797.pdf (accessed on 18 September 2017). 23 Hans-Christoph Burmeister et al., Autonomous Unmanned Merchant Vessel and its Contribution towards the e-Navigation Implementation: The MUNIN Perspective, International Journal of eNavigation and Maritime Economy 1 (2014), 1.

AUTOMATED COMMERCIAL VESSELS

117

board/off-board officer of navigation decisions), and full autonomy (full situation awareness, assessment, and decision-making without human intervention).24 The early legal discourse on MASS raised the question whether the autonomous vessel is in fact a ‘ship’ as generally understood in international maritime law in order to ascertain its legal status and application of the various public and private maritime law conventions.25 While ‘ship’ and ‘vessel’ are used interchangeably in the United Nations Convention on the Law of the Sea (UNCLOS),26 they are not defined in that instrument. Rather, the search for definitions leads to the international maritime conventions, several of which are considered below, but without attempting an exhaustive listing. In most cases the term ‘ship’ is defined and the definition includes ‘vessel(s)’ having specific characteristics and/or performing particular functions for the purpose of delimiting the scope and application of the instrument concerned. A perusal of several of these instruments illustrates the drafting practice. In the marine pollution conventions, ‘ship’ is defined as or includes: “any sea-going vessel of any type whatsoever”;27 “a vessel of any type whatsoever operating in the marine environment and includes hydrofoil boats, air-cushion vehicles, submersibles, floating craft and fixed or floating platforms”;28 “any sea-going vessel and seaborne craft of any type whatsoever constructed or adapted for the carriage of oil in bulk as cargo […]”;29 “a vessel of any type whatsoever operating in the marine environment and includes hydrofoil boats, air-cushion vehicles, submersibles, and floating craft of any type”;30 24 Mogens Blanke/Michael Henriques/Jakob Bang, A Pre-Analysis on Autonomous Ships (2016), 3, available at: https://www.dma.dk/Documents/Publikationer/Autonome%20skibe_DTU_rapport_UK. pdf (accessed on 18 September 2017). 25

For example see Comité Maritime International (CMI), CMI International Working Group Position Paper on Unmanned Ships and the International Regulatory Framework, 2017, 3, available at: http://www.comitemaritime.org/Uploads/Questionnaires/CMI%20Position%20Paper%20on%20 Unmanned%20Ships.pdf (accessed on 18 September 2017). 26 United Nations Convention on the Law of the Sea, 10 December 1982, UNTS 1833, 396 (UNCLOS). 27 Art. II (2) International Convention Relating to Intervention on the High Seas in Cases of Oil Pollution Casualties, 29 November 1969, UNTS 970, 211; Art. 1 (1) International Convention on Civil Liability for Bunker Oil Pollution Damage, 23 March 2001, IMO Doc. LEG/CONF.12/19 (2001); Art. 1 (1) International Convention on Liability and Compensation for Damage in Connection with the Carriage of Hazardous and Noxious Substances by Sea, 2 May 1996, ILM 35, 1406. 28

Art. 2 (4) MARPOL.

29

Art. I (1) CLC Convention. An identical definition is included in Art. I (1) IOPCF Convention.

30

Art. 2 (3) International Convention on Oil Pollution Preparedness, Response and Cooperation, 30 November 1990, UNTS 1891, 51.

118

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

a vessel of any type whatsoever operating in the marine environment and includes hydrofoil boats, air-cushion vehicles, submersibles, floating craft, fixed or floating platforms, floating storage units (FSUs) and floating production storage and off-loading units (FPSOs)”;31 and “a vessel of any type whatsoever operating in the aquatic environment and includes submersibles, floating craft, floating platforms, FSUs and FPSOs”.32 In a maritime security convention, ‘ship’ is defined as “a vessel of any type whatsoever not permanently attached to the sea-bed, including dynamically supported craft, submersibles, or any floating craft.”33 The definitions of ‘ship’ for maritime carriage purposes include: “vessel used for the carriage of goods by sea”;34 “only a seagoing vessel, excluding air-cushion vehicle”;35 “any self-propelled sea-going vessel used in international seaborne trade for the transport of goods, passengers, or both with the exception of less than 500 gross registered tons”.36 They share the common themes of ‘sea-going’ and/or performing marine transportation. It is unusual for ‘ship’ and ‘vessel’ to be inverted, as in the case of a salvage convention where “vessel means any ship or craft or any structure capable of navigation”37 and collision avoidance rules where ‘vessel’ is defined to include “every description of water craft, including non-displacement craft and seaplanes, used or capable of being used as a means of transport on water.”38 However, for the purpose of the discussion of autonomous ships the difference is immaterial and none of the definitions provided above pose a problem for the consideration of autonomous vessels as ‘ships’. Accordingly, the ship will be subject to domestic and international safety, environmental, and security regulations. The rules and standards will govern its entire life 31 Art. 2 (9) International Convention on the Control of Harmful Anti-Fouling Systems on Ships, 5 October 2001, IMO Doc. AFS/CONF/26 Annex (2001) (AFS Convention). 32 Art. 1 (12) International Convention for the Control and Management of Ships’ Ballast Water and Sediments, 13 February 2004, IMO Doc. BWM/CONF/36 (2004) (BWM Convention). 33 Art. 1 Convention for the Suppression of Unlawful Acts against the Safety of Maritime Navigation, 10 March 1988, UNTS 1678, 201. 34 Art. 2 International Convention for the Unification of Certain Rules Relating to the Limitation of Liability of Owners of Seagoing Vessels, 25 August 1924, LNTS 120, 123. 35

Art. 1 (3) Athens Convention Relating to the Carriage of Passengers and their Luggage by Sea, 13 December 1974, UNTS 1463, 20. 36

Art. 2 United Nations Convention on Conditions for Registration of Ships, 7 February 1986, ILM 26, 1229. 37

Art. 1 (b) International Convention on Salvage, 28 April 1989, UNTS 1953, 165.

38

Rule 3 (a) COLREGS.

AUTOMATED COMMERCIAL VESSELS

119

from construction, through operations, and eventual recycling. Its operational life will include rules with respect to its seaworthiness and crewing. The private law consequences include the generation of liens unique to ships, the ability of a creditor to arrest and proceed in rem against the ship, and the ability of the owner, among other related interests in the ship, to enjoy certain defences, most especially limitation of liability. The ship is a piece of property that enjoys a status unique to maritime law. The early conceptions and prototypes of MASS clearly foresee vessels that perform the full range of functions of ships that are fully manned. The distinctive characteristics are the absence of a crew and in the case of a fully autonomous vessel, the relegation of decision-making to an automated system. While factoring the withdrawal of the human element, there is not much reason not to consider MASS as capable to produce most of the legal consequences usually associated with ships. Rather than with definitional concerns, the focus should be on the impact of the absence of or lessened human element on particular institutions that are premised on human appraisal, judgement, and decision-making.

III. Implications for the International Law of the Sea MASS are likely to raise some questions of interpretation and application of the conventional and customary law of the sea, in particular the UNCLOS of 1982. The key question posed is the extent to which the pertinent UNCLOS regimes are sufficiently broad so as to seamlessly support the advent of MASS, or whether the characteristics of these new technologies require revisiting of the assumptions and expected applications of particular rules. Of special interest are regimes where MASS enjoy navigational rights as all other ships and jurisdictional powers with respect to international shipping based on the assumption that ships are manned, more specifically with respect to flag, coastal, and port States. At first blush, the frequent reference to the generic terms ‘vessel’ or ‘ship’ in the UNCLOS suggests that the provisions concerned apply to all ships including MASS. However, a probing analysis of some provisions raises potential issues.39

39

MSC (note 17), 79.

120

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017 A. Flag State

The flag State has a right to register any class of ship and by doing so will be extending the application of its laws and protections to the ship wherever it navigates.40 The question as to whether a national register of shipping is able to accommodate MASS is not an issue in the UNCLOS and indeed for many flag States that do not require manning as a condition for the registration of a ship. Registration requirements usually include ship name, IMO number, ownership, technical particulars, whether the vessel is on a bareboat charter, whether the vessel’s previous registration is terminated, and also mortgages or hypothecs on the ship and their status. If registration of MASS ships is an issue for some States, it is likely because of domestic requirements. Once a vessel is entered into the register of ships, the flag State commences to enjoy exclusive jurisdiction over those vessels when they are on the high seas,41 concurrent jurisdiction with coastal States during the exercise of innocent passage,42 transit passage,43 archipelagic sea lanes, and innocent passage in archipelagic waters,44 and concurrent jurisdiction with port States when ships are voluntarily in a foreign port.45 The flag State has a right to provide them with consular assistance in foreign ports.46 The right to establish a ship register is accompanied by the duty to effectively exercise jurisdiction and control over MASS entered into the register.47 The actions usually associated with this power would need to be reconsidered in the case of unmanned ships. If the ship does not carry a crew the exercise of jurisdiction and control will be expected to focus on administrative and technical, rather than the maritime labour aspects (master, officers, and crew) as would be the case with crewed ships.48 As the technology of MASS develops, it is conceivable that new rules (international and domestic) governing the desired competencies and capabilities of the shore-board 40

Arts. 91–92 UNCLOS.

41

Art. 92 (1) UNCLOS.

42

Art. 21 UNCLOS.

43

Arts. 42 and 45 UNCLOS.

44

Arts. 49, 52–54 UNCLOS.

45

Art. 8 (1) UNCLOS.

46

Arts. 27 and 231 UNCLOS.

47

Art. 94 (1) UNCLOS.

48

Art. 94 (1) and (2) UNCLOS.

AUTOMATED COMMERCIAL VESSELS

121

personnel responsible as well as the instrumentation for the remote or automated navigation of the ship, including communications, will be needed.49 More of an issue in the UNCLOS is the duty of the flag State to take such measures for ships flying its flag as are necessary to ensure safety at sea with regard, inter alia, to: (a) the construction, equipment and seaworthiness of ships; (b) the manning of ships, labour conditions and the training of crews, taking into account the applicable international instruments; (c) the use of signals, the maintenance of communications and the prevention of collisions.50

The contemplated measures include requirements concerning technology of the traditional ship used by an on-board crew, rather than MASS. For example, the measures are expected to include on-board “charts, nautical publications and navigational equipment and instruments as are appropriate for the safe navigation of the ship.”51 While a fully autonomous ship will have navigation equipment, it may not have charts on board, but rather will be able to download chart and navigational data as pre-programmed and as needed. Admittedly, electronic charts, digital manuals useable online, and electronic notices to shipping are now already in widespread use. A more difficult requirement for the flag State is to ensure that each ship is in the charge of a master and officers who possess appropriate qualifications, in particular in seamanship, navigation, communications and marine engineering, and that the crew is appropriate in qualification and numbers for the type, size, machinery and equipment of the ship.52

Simply stated, the rule requires the ship to have qualified personnel on board. One could argue that the crew requirement in this rule is relative to the type of ship and machinery, and if the ship is partially or fully automated, then it follows that a crew is not needed all of the time, if at all, without violating the rule. Thus, the practise of using towed, but unmanned barges for carriage of goods is of long standing. Further, if the fundamental purpose of the manning requirement is to ensure that there is competent operation and navigation of the vessel, it could be argued that a MASS vessel could be competently operated and navigated in a remote or autonomous man49 CMI (note 25), 16. This paper observes that unmanned ships will introduce a new range of personnel and qualification standards. 50

Art. 94 (3) UNCLOS.

51

Art. 94 (4)(a) UNCLOS.

52

Art. 94 (4)(b) UNCLOS.

122

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

ner where it can be demonstrated that safety, environmental protection, and security are assured at least to the same extent as with a manned vessel. For remotely operated vessels, “in the charge of” could be constructively interpreted to include the shore-based team of professionals in charge of the vessel’s navigation, but presumably the standard of human competence still applies to the shore-based team. Arguing that the element of human control is nonetheless manifested in the writer who creates the algorithm appears to stretch too far because the autonomous ship is expected to operate with ongoing human control. A similar interpretation issue arises with respect to remotely controlled and autonomous warships. The definition of ‘warships’, while specifically requiring that they are under the armed forces of a State, further specifies that they are “under the command of an officer duly commissioned by the government of the State and whose name appears in the appropriate service list or its equivalent, and manned by a crew which is under regular armed forces discipline.”53 However, warships are excepted from the mandatory application of the International Convention for the Safety of Life at Sea of 1974 (SOLAS Convention)54 and the International Convention on Standards of Training, Certification and Watchkeeping for Seafarers of 1978 (STCW Convention).55 The rule that the master, officers and, to the extent appropriate, the crew are fully conversant with and required to observe the applicable international regulations concerning the safety of life at sea, the prevention of collisions, the prevention, reduction and control of marine pollution, and the maintenance of communications by radio56

is a challenge for ships in autonomous mode and will be considered further below. For remotely controlled ships, and as discussed in the previous paragraph, the shorebased team could be argued to provide the human constructive presence and expertise 53

Art. 29 UNCLOS.

54

Ch. 1, Reg. 3 (a) SOLAS Convention.

55

Art. 3 (a) International Convention on Standards of Training, Certification and Watchkeeping for Seafarers, 7 July 1978, UNTS 1361, 2 (STCW Convention). The exception applies to “(a) warships, naval auxiliaries or other ships owned or operated by a State and engaged only on governmental noncommercial service; however, each Party shall ensure by the adoption of appropriate measures not impairing the operations or operational capabilities of such ships owned or operated by it, that the persons serving on board such ships meet the requirements of the Convention so far as is reasonable and practicable”. 56

Art. 94 (4)(c) UNCLOS.

AUTOMATED COMMERCIAL VESSELS

123

underlying this requirement. However, for autonomous ships it is difficult to argue that the algorithm substitutes the requirement for fully-conversant master, officers, and crew in this provision. As for the communications aspect, which entails an ability for other vessels and public authorities to communicate with the vessel, the shorebased personnel behind a remotely controlled ship could conceivably perform this function. This constructive interpretation of the UNCLOS provisions does not easily extend to vessels in autonomous mode because decision-making is left to algorithms, rather than to direct human decision-making, and two-way communication involving a machine on the autonomous vessel and humans on vessels in its vicinity (for instance using very high frequency radio (VHF) to avoid close quarters) may not be possible, let alone desirable. In performing these duties, the flag State “is required to conform to generally accepted international regulations, procedures and practices and to take any steps which may be necessary to secure their observance.”57 This provision invokes international regulations that may only be adopted by the IMO and at this time the pertinent maritime conventions do not directly address the particular requirements and perhaps exemptions (for example from manning requirements) that will be needed for MASS ships. Flag States are required to ensure compliance of their ships with international rules and standards and provide for effective enforcement.58 This duty extends to preventing their ships from operating unless they comply with requirements for design, construction, and manning,59 and to require them to carry on board certificates which may be required under various conventions,60 such as SOLAS61 and the International Convention for the Prevention of Pollution from Ships of 1973/78 (MARPOL).62 While some documentation is already available in digital format and there is a trend towards passive interactive ship-shore reporting, the operationalisation of the rule concerning documentation on board MASS will need to be reconsidered when an

57

Art. 94 (5) UNCLOS.

58

Art. 217 (1) UNCLOS.

59

Art. 217 (2) UNCLOS.

60

Art. 217 (3) UNCLOS.

61

For example Ch. I, Reg. 16 SOLAS Convention on availability of certificates on board the ship.

62

For example Annex I, Reg. 10 MARPOL with respect to the International Oil Pollution Certificate.

124

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

inspecting coastal or port State authority official requests the certificates, as will be seen below with respect to port State control. Another constraint with respect to MASS is the flag State’s responsibility under the UNCLOS to require the master of a ship to offer assistance to persons in distress at sea,63 also a duty under SOLAS.64 This duty includes assistance to persons, deviating from the established course to proceed with speed to the rescue of persons in distress, and rendering assistance to the other ship involved in a collision. It is difficult to see how an unmanned ship can discharge this humanitarian duty, other than, perhaps, for the shore-based team to notify rescue services in the area of the incident and leave the MASS vessel in the area just in case search and rescue response services need to use the vessel as a platform, if appropriate. In summary, the rules concerning duties of the flag State in the UNCLOS reflect a combination of the need for constructive interpretation of some provisions and addressing regulatory gaps with respect to ships that are unmanned. They potentially pose a problem for the flag State in the exercise of effective jurisdiction and control if the duty to ensure the manning of ships is interpreted literally as a requirement to have crew members on board of the MASS. However, if the deployment of a temporary crew for port approach navigation, berthing, and de-berthing, and a shore-based team for the long stretches of navigation of the vessel are interpreted constructively to be the equivalent of the actual manning in the case of remotely controlled ships, the gap is effectively removed. The same cannot be said for fully autonomous ships if these indeed become a reality, thus leaving a gap that would need to be addressed through other means.

B. Coastal State

The coastal State’s limited jurisdiction over ships exercising the rights of innocent passage through the territorial sea, sea lanes passage through archipelagic waters, and transit passage through straits used for international navigation extends to all ships, but not to warships.65 The coastal State may adopt laws and regulations to govern in63

Art. 98 UNCLOS.

64

Ch. V, Reg. 33 SOLAS Convention.

65

Art. 32 UNCLOS.

AUTOMATED COMMERCIAL VESSELS

125

nocent passage in the territorial sea (and archipelagic waters)66 with respect to safety of navigation,67 and in doing so must not apply “design, construction, manning or equipment [standards] unless they are giving effect to generally accepted international rules or standards.”68 At this time, neither the UNCLOS (supra) nor SOLAS (infra) exempts ships from actual manning requirements. In theory, a coastal State could regulate a manning requirement for MASS in transit, especially in areas of heavy traffic, because in doing so it would not be applying a standard other than a generally accepted international manning standard. An analogous argument could be made with respect to laws and regulations for pollution prevention.69 With respect to transit passage, the coastal State has less flexibility in finding issues with MASS because its laws and regulations must “not discriminate in form or in fact among foreign ships or in their application have the practical effect of denying, hampering or impairing the rights of passage […]”70 The UNCLOS text in this part does not provide the coastal State with a justification to impose a manning requirement. There are potential difficulties that coastal States could face in undertaking enforcement action in accordance with UNCLOS. In undertaking enforcement action against MASS that violate pollution prevention laws applicable in the territorial sea or exclusive economic zone (EEZ) when they are voluntarily in its port, the coastal State has to contend with a scenario where it is unclear whether there is a human person behind the violation and, if there is, how to exercise jurisdiction over a human person (or corporation) who may not be within its territorial jurisdiction, despite the vessel’s voluntary presence.71 The coastal State has the power to undertake physical inspection of MASS, and where the evidence so warrants, it may detain such vessels although there are no persons on board.72 It could be difficult for the coastal State to pin responsibility to a specific person. A commonly applied practical approach to pollution offences is strict liability, so that the actus reus alone will be sufficient to proceed against the ship. Another challenge is how a coastal State authority is to commu66

Art. 52 (1) UNCLOS.

67

Art. 21 (1)(a) UNCLOS.

68

Art. 21 (2) UNCLOS.

69

Art. 211 (6) UNCLOS.

70

Art. 42 (2) UNCLOS.

71

Art. 220 (1) UNCLOS.

72

Art. 220 (2) and (6) UNCLOS.

126

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

nicate with MASS when unmanned. When it has clear grounds to believe that there is a pollution incident, the coastal State may require such a vessel to provide “information on its identity and port of registry, its last and its next port of call and other relevant information required to establish whether a violation has occurred.”73 While communication with a remotely operated vessel should not be an issue because there will be contact with a shore-based crew, two-way communication with a vessel in autonomous mode could be difficult when inspectors need to conduct a lengthy exchange unless a shore-based team intervenes. The power of a coastal State to conduct a physical inspection could be constrained.74 The first stage of physical inspection is limited to an examination of on-board certificates, records, or other documents,75 which as seen earlier may simply be digital. A second and more probing inspection stage may be undertaken if there are issues with such documents, and where it is established that a violation has occurred, the vessel has a right to release after reasonable procedures are completed, such as posting of a bond or other appropriate financial security, just like any other ship.76 The coastal State may refuse release only exceptionally or release conditionally, notifying the flag State in such eventualities, who in turn may still seek prompt release.77 A potentially more complex issue for a coastal State is a scenario where MASS are in need of assistance and request refuge in a port or safe waters of that State. The custom of providing refuge to ships in distress has been narrowed down in State practice to humanitarian assistance.78 An unmanned ship does not need humanitarian assistance, but its cargo could be very valuable and the vessel’s distressed condition could be such as to threaten the marine environment. Although the coastal State has rights of self-protection under conventional law, it will be guided by the IMO Guidelines for Ships in Need of Assistance.79 While the guidelines are clearly voluntary, they are increasingly considered as good practice for risk-based decision-making. However, the 73

Art. 220 (3) UNCLOS.

74

Art. 220 (5) UNCLOS.

75

Art. 226 (1)(a) UNCLOS.

76

Art. 226 (1)(b) UNCLOS.

77

Arts. 226 (1)(c) and 292 UNCLOS.

78

ACT Shipping (OTE) Ltd. v. Minister for the Marine, Ireland and the Attorney-General (The MV Toledo) [1995] 2 ILRM 30. 79

Guidelines on Places of Refuge for Ships in Need of Assistance, IMO Assembly Resolution A.949(23) of 5 December 2003, IMO Doc. A 23/Res.949 (2004).

AUTOMATED COMMERCIAL VESSELS

127

problem with the guidelines is that they assume that the vessel in need of assistance has a master and crew on board, and that usually the vessel will be receiving salvage assistance as well. The guidelines establish a communication process between the three parties to enable coastal State decision-makers to decide on refuge. Usually, the master and crew will be the persons who have first and most in-depth knowledge of the condition of the ship. The absence of such persons on board the vessel could constrain the decision-making process on refuge.

C. Port State

The port State’s concern with MASS is with respect to the enforcement of international rules and standards and responding to the requests of other States concerning the autonomous vessel ‘voluntarily’ in its port.80 The notice of arrival, which can be verbal or electronic, constitutes an indication of the intention to enter port. When a vessel is selected for inspection, there is usually communication between the port State control inspector and the ship master. Presumably, such communication may be undertaken with the shore-based team, but the inspector will need to board the ship. Usually, the inspector will board and personally communicate with the master or officer on watch (OOW). If the autonomous vessel does not have personnel on board in the foreign port, it is conceivable that the ship’s agent, usually appointed to act on behalf of its interests in a foreign port, may need to step in. However, the ship agent is usually a commercial, rather than a person with technical knowledge of the ship as a crew member. Thus the inspector’s questions may need to be directed to persons who are not present on the ship during the inspection, or simply wait until such persons are available. When an autonomous vessel is port-inspected and found to have discharged wastes in violation of international standards, there is the further enforcement constraint with respect to who to proceed against when the ‘person’ responsible for the discharge is outside of the territorial jurisdiction of the port State. The UNCLOS does not indicate what persons would be pursued in the institution of proceedings and the vessel is personified in the sense that it is expected to comply with the inspection.81 80

Art. 218 UNCLOS.

81

Art. 218 (3) UNCLOS.

128

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

The port State has the power to detain the vessel to persuade the person(s) concerned to submit to jurisdiction. The port State would need to inform the flag State.

IV. Implications for International Maritime Law The article now considers the potential impact of MASS on the international maritime legal system. Consisting of conventions and other legal instruments of the IMO, international maritime regulation addresses all ships and is largely premised on a human presence on board and in control of the navigation of the ship, as well as performing contingent functions such as provision of notices, log book entries, reporting, and responding to distress calls for assistance. The provision of certain services to ships is also premised on human interaction, such as in the case of pilotage where the master and OOW cooperate with the pilot in formulating the passage plan and taking and executing navigation instructions. The human element is underscored by trained crews. For the purposes of the present discussion, a selection of salient instruments and provisions are discussed to exemplify the potential impacts of MASS on contemporary maritime regulation, while acknowledging that a comprehensive analysis of all IMO regulations will be needed to ascertain the full regulatory impact of MASS.

A. Maritime Safety

The SOLAS Convention constitutes the backbone of international maritime safety regulation. Through fourteen chapters and numerous codes, it sets out standards, inter alia, for ship construction, machinery, cargo handling, stability, life-saving equipment and procedures, radiocommunications, safe carriage of cargoes and oil fuels, carriage of dangerous goods, additional standards for specific types of vessels (e.g. nuclear ships, bulk carriers, high speed craft), safe navigation, management for safe ship operations, maritime security, and polar shipping. In principle, and whatever function they perform, MASS will be captured by a range of SOLAS rules and standards. SOLAS establishes a comprehensive system for the inspection, surveying, and certification requirements for ships to ensure compliance with international stan-

AUTOMATED COMMERCIAL VESSELS

129

dards.82 The principal issues that MASS raise are with respect to safety aspects where the human factor is concerned. As observed in the earlier discussion with respect to the flag State’s duties under the UNCLOS, a key SOLAS rule provides that “Contracting Governments undertake, each for its national ships, to maintain, or, if it is necessary, to adopt measures for the purposes of ensuring that, from the point of view of safety of life at sea, all ships shall be sufficiently and efficiently manned”.83 The intention of this rule is to ensure that all ships are sufficiently manned to ensure they can operate safely. Crewing is equated with the safety outcome. An interpretation of this provision analogous to the earlier discussion on a similar requirement in Article 94 (3) UNCLOS is that as long as safety outcomes at least comparable to, if not exceeding, manned vessels, can be achieved by MASS vessels, the intended outcome of the rule is achieved. The presence of a shore-based team in remote control or standby (or an on-board crew on standby) and temporary crews on unmanned autonomous ships provides the overseeing human element implicit in this rule. Naturally, to remove any doubt, it would be preferable for this manning rule to be amended to reflect the needs of MASS. The application of SOLAS principles and standards relating to bridge design and location of equipment are premised on personnel manning the bridge.84 The ergonomics should be such as to facilitate the “tasks to be performed by the bridge team and the pilot in making full appraisal of the situation and in navigating the ship safely under all operational conditions” and “safe and effective bridge resource management.”85 As observed earlier, MASS will be guided by a different bridge concept, possibly much smaller for fully autonomous vessels in the event a crew is needed on board, and with a shore-based equivalent. This is important from another perspective. In gathering information on navigational conditions, the ship is also in a position, and indeed has a duty, to relay information on any dangerous conditions encountered. Thus the master has a duty to communicate to ships in the vicinity and competent authorities any danger messages concerning the presence of dangerous ice, dangerous derelict, other direct danger to navigation, tropical storms, encounters with sub-freezing air temperatures associated with gales, and winds of force 10 or above on the Beau82

Ch. I SOLAS Convention.

83

Ch. V, Reg. 14 SOLAS Convention.

84

Ch. V, Reg. 15 SOLAS Convention.

85

Ibid.

130

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

fort scale not forecasted.86 At this time it is unclear how MASS would comply with the danger message rule. It is likely that the new international standards would need to be developed and regulated for the new instrumentation and processes that will replace these human tasks. At this time SOLAS does not have rules addressing the shore-based control room. In remotely controlled and autonomous vessels, the bridge and watch tasks performed by humans will be replaced by a suite of sensors and daytime and infrared cameras in addition to the usual instrumentation, such as radar and AIS, to enable processing of complex data streams and enable full appraisal of the ship’s location and situation. Rules concerning the maintenance of a watch during the operation of the vessel are currently premised on crew presence. While at sea a ship is required to maintain a continuous watch, including a radio watch for broadcasts of maritime safety information.87 For watch purposes, SOLAS requires that every ship shall carry personnel qualified for distress and safety radiocommunication.88 The pertinent rules would need to be adapted to enable performance of watch functions by the off-board team responsible for the ship or to permit automated task performance. With no crew, the vessel might not need the usual number of survival craft. The rules concerning the manning of survival craft would have contingent (in the event the vessel is boarded), if any application. The general rule for all ships is that they have a sufficient number of trained persons on board for mustering and assisting untrained persons, and a sufficient number of crew members must be available for launching an on-board survival craft.89 MASS could justify an exemption from or modification of these rules. The International Convention on Load Lines of 1966 (LLC),90 another vital maritime safety instrument, in addition to setting out rules for the safe loading of ships, further provides for construction standards to ensure safe work and passage spaces on board. The LLC has provisions for the protection of the crew, such as standards for 86

Ch. V, Reg. 31 SOLAS Convention.

87

Ch. IV, Reg. 12 SOLAS Convention.

88

Ch. IV, Reg. 16 SOLAS Convention.

89

Ch. III, Reg. 10 SOLAS Convention.

90

International Convention on Load Lines, 5 April 1966, UNTS 640, 133 (LLC); Protocol of 1988 relating to the International Convention on Load Lines, 11 November 1988, IMO Doc. HSSC/ CONF/12 (1988).

AUTOMATED COMMERCIAL VESSELS

131

deckhouses for accommodation, guard rails, safe passageways, and stowage of cargo in a manner to permit safe crew passage,91 which might not be required in the design and construction of MASS to the same extent. It is likely that an exemption for MASS will be also be required in this instrument.

B. Crew Training, Certification, and Work Conditions

The International Convention on Standards of Training, Certification and Watchkeeping for Seafarers of 1978 (STCW Convention) is the principal international maritime instrument dedicated to the training and certification of seafarers. The STCW Convention establishes that “each Administration shall require every company to ensure that its ships are manned in compliance with the applicable safe manning requirements of the Administration”.92 Shipping companies93 are responsible for the manning of their ships in compliance with the Convention and that properly certificated seafarers are assigned for service on their ships.94 The ship’s complement must be able to “effectively co-ordinate their activities in an emergency situation and in performing functions vital to safety or to the prevention or mitigation of pollution.”95 Levels of safe manning are further addressed by the Principles of Minimum Safe Manning.96 In theory, where MASS are fully autonomous, they will render redundant the application of much of the STCW Convention, as well as the important Maritime La-

91

Annex I, Ch. II, Regs. 25 and 25-1 LLC.

92

Ch. I, Reg. I/14 (1) STCW Convention.

93

‘Company’ is defined as “the owner of the ship or any other organization or person such as the manager, or the bareboat charterer, who has assumed the responsibility for operation of the ship from the shipowner and who, on assuming such responsibility, has agreed to take over all the duties and responsibilities imposed on the company by these regulations.”, Ch. I, Reg. I/1 (23) STCW Convention. 94

Ibid.

95

Ibid.

96

IMO Assembly, Principles of Minimum Safe Manning, Resolution A.1047(27) of 30 November 2011, IMO Doc. A 27/Res.1047 (2011). The Principles consist of: Annex 1 – Guidelines for the Application of Principles of Safe Manning; Annex 2 – Guidelines for Determination of Minimum Safe Manning; Annex 3 – Responsibilities in the Application of Principles of Minimum Safe Manning; Annex 4 – Guidance on Contents and Model Form of Minimum Safe Manning Document; and Annex 5 – Framework for Determining Minimum Safe Manning.

132

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

bour Convention of 2006 (MLC)97 which promotes the rights of seafarers and establishes a framework for conditions of employment and occupational health and safety on ships. Where MASS have crews deployed temporarily, the STCW Convention and MLC will apply. The MLC protections are likely not extendable to the team operating MASS from shore as the definition of seafarer presumes work at sea.98 In practice, there are a number of vital functions performed on board the ship, both in normal and emergency situations pursuant to SOLAS and articulated in terms of personnel requirements, that would need to be addressed by MASS and the teams responsible for their operation or for overseeing the systems that run the vessels. The potential impact of automation is already recognised in the framework for determining minimum safe manning, in particular that functions on board a ship “may increase or decrease manning levels depending on availability and appropriate procedures and of specific capability enabling technology/automation.”99 The usual personnel training requirements reflect the various functions on board the ship. For deck operations, the rules include mandatory minimum requirements for officers in charge of the navigational watch on ships of 500 gross tonnage (GRT) or more100 as well as for certification of masters and chief mates for ships of both more and less than 500 GRT,101 and certification of ratings.102 There are similar requirements for the engine room department. The rules include certification on the engineering watch both in manned and, when on duty, in unmanned engine rooms,103 and mandatory minimum requirements for certification of ratings on the watch in manned engine-rooms and, when periodically on duty, in unmanned engine-rooms.104 The recent updating of standards to reflect the reality that ships may have unmanned engine rooms for prolonged periods is evidence that certain ship functions are already capable of full automation. Requirements for radiocommunications and radio personnel include mandatory minimum requirements for certification for Global Maritime 97

Maritime Labour Convention, 23 February 2006, ILM 45, 792 (MLC).

98

According to Art. 2 (1)(f) MLC ‘seafarer’ means “any person who is employed or engaged or works in any capacity on board a ship to which this Convention applies”. 99

Principles of Minimum Safe Manning (note 96), Annex 5 at 1.5.

100

Ch. II, Reg. II/1 STCW Convention.

101

Ch. II, Reg. II/2-3 STCW Convention.

102

Ch. II, Reg. II/4 STCW Convention.

103

Ch. III, Reg. III/1 STCW Convention.

104

Ch. III, Reg. III/4 STCW Convention.

AUTOMATED COMMERCIAL VESSELS

133

Distress and Safety System (GMDSS).105 There are also mandatory minimum special personnel training requirements for certain classes of ships, such as tankers, ro-ro ships, and passenger vessels.106 A vital rule in the Convention is that “a safe continuous watch or watches appropriate to the prevailing circumstances and conditions are maintained in all seagoing ships at all times”107 under the direction of the master. The interpretation of ‘watch’ functions with respect to SOLAS are relevant here as well. The STCW Convention effectively defines the standards of competence expected of all human-performed functions on board at all times. The Convention ensures human competence in decision-making. While MASS will largely shed off the human component and integrate the various functions into a single system, they are still expected to produce comparable safety, environmental, and security outcomes as manned ships. During the voyage, the system, whether overseen remotely from offboard and/or in an automated manner, provides centralised overview and management of all ship operations. While in port, MASS would be serviced and maintained as other ships. With respect to shore-based personnel monitoring and/or in control of MASS, ‘competence’ will continue to be a requirement, although the precise content will need to be redefined and adapted. The shore-based personnel will be highly qualified personnel in navigational, engineering, and informatics aspects of the vessel. Standards for their training and certification will likely need to be enhanced. In the case of fully automated vessels, autonomous decision-making introduces a novel approach to ‘competence’ and it is likely that new technical standards will need to be developed. Also, and as observed earlier, the STCW rules on manning requirements do not apply to warships. The exception applies to “warships, naval auxiliaries and other ships owned or operated by a State and engaged only on governmental non-commercial service.”108 This exception is accompanied by an important proviso to the effect that

105

Ch. IV, Reg. IV/2 STCW Convention.

106

Ch. V STCW Convention.

107

Ch. VIII, Reg. VIII/1 STCW Convention.

108

Art. III (a) STCW Convention.

134

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

each Party shall ensure by the adoption of appropriate measures not impairing the operations or operational capabilities of such ships owned or operated by it, that the persons serving on board such ships meet the requirements of the Convention so far as is reasonable and practicable.109

C. The ‘Rules of the Road’

The ‘rules of the road’ for the safe navigation of ships are set out in the Convention on the International Regulations for Preventing Collisions at Sea of 1972 (COLREGS) and apply to the high seas and all navigable waters.110 The COLREGS include core rules for safe navigation (Part A – general; Part B – steering and sailing rules) in conditions of normal and reduced visibility and signalling to other ships in the vicinity (Part C – lights and shapes; Part D – sound and light signals). The rules nourish the standard of good seamanship expected of the ordinary mariner and against which the standard of care in the navigation of a vessel is assessed. Thus the rules provide no basis for exemptions from compliance with the rules or from exercising the precaution necessary and expected in the ordinary practice of seamen.111 Although unmanned, MASS are required to observe the COLREGS and the underlying standard of good seamanship. The rules are explicit in their reference to vessels and therefore do not provide any immediate problem of textual interpretation. The definition of vessel as “every description of water craft” used or capable of being used as means of transportation over water easily covers MASS.112 MASS will not be expected to enjoy special privileges. They do not qualify as vessels ‘not under command’, which enjoy the privilege of a stand-on vessel, because this technical term refers only to “a vessel which through some exceptional circumstance is unable to manoeuvre as required by these Rules and is therefore unable to keep out of the way of another vessel.”113 Rather, the issues that could arise likely stem from how they would be applied when an autonomous vessel is in the vicinity of other vessels and the potential legal consequences. 109

Ibid.

110

Rule 1 (a) COLREGS.

111

Rule 2 (a) COLREGS.

112

Rule 3 (a) COLREGS.

113

Rule 3 (f) COLREGS.

AUTOMATED COMMERCIAL VESSELS

135

In general, the Part B rules apply to MASS without significant issues of interpretation. The steering rules require that every vessel must maintain a proper look-out in any condition of visibility. This is described as “by sight and hearing as well as by all available means appropriate in the prevailing circumstances and conditions so as to make a full appraisal of the situation and of the risk of collision.”114 As described earlier, MASS will be equipped with a sophisticated system of technologies for full situational awareness, including geopositioning, radar, sensors, and continuous streaming cameras, all generating data for an algorithm to translate into information and advice for human or automated decision-making. In theory, automation will be able to more accurately navigate the vessel and avoid the human element contribution to maritime collisions and allisions. There is a lesson to be learned from over-reliance on technology for safe navigation. Following the introduction of radar on board ships, there were numerous radar-assisted collisions that were ultimately attributed to poor lookout. Seafarers at the time misused radar or over-relied on its use, thus refraining from using all means available to determine the position of other vessels and communicate with them if necessary to avoid close quarters.115 There could also be complex legal consequences in determining contributory negligence and consequential liabilities, in particular where the vessel is fully automated. In collision cases, unlike allisions, it is often the case that the collision is the result of negligence on the part of both ships involved. Another rule whose operationalisation will need careful consideration concerns the risk of collision and the action to be taken to avoid close quarters. The vessel has the duty to “use all available means appropriate to the prevailing circumstances and conditions to determine if risk of collision exists” and in case of doubt to deem that a risk of collision exists.116 The rule is supported by further rules concerning the proper use of radar (learning from radar-assisted collisions) and avoiding of assumptions on scanty information. This rule relies heavily on data inputs on traffic within the vicinity of the vessel, general conditions of navigation (e.g. weather, visibility, ice presence, etc.), proper use of equipment, and intelligent judgement. It is argued that MASS will

114

Rule 5 COLREGS.

115

Nicholas J. Healy, Radar and the New Collision Regulations, Tulane Law Review 37 (1962–1963),

116

Rule 7 (a) COLREGS.

37.

136

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

be better able to collect data and make intelligent decisions.117 In less than optimal conditions when information is scanty, the extent to which human experience and judgement call, accompanied by good seamanship, can be replaced by an algorithm remains to be seen. The Part C rules on lights and shapes, respectively for night time and day time sailing, will apply. These rules are important to enable vessels in each other’s vicinity to recognise their respective status and privilege and apply the respective rule for standon and give-way vessels while ensuring safe distances. While not granting them any special privilege, the introduction of dedicated new lights and shapes for MASS ships to indicate their presence to other vessels in their vicinity could be a sensible and precautionary addition to the rules. The Part D rules on sounds and signals also apply, although the actual operation of manoeuvring and warning whistle blasts and light signals would be done remotely or automatically.

D. Environment Protection

The absence of crews on MASS can be expected to produce impacts on environmental regulation with respect to on-board procedures, waste management, and reporting requirements. Again, examples are provided to highlight the point. MARPOL Annex I on-board requirements for the master’s or crew’s familiarity with essential shipboard procedures relating to the prevention of pollution by oil in the case of unmanned ships would need to be reviewed to enable a port State inspector to ascertain compliance with requirements in some other manner.118 The absence of a crew might significantly reduce the relevance of Annexes IV and V respectively concerning sewage and garbage. There would be a strong argument to consider the granting of exemptions from certain certification requirements with respect to these wastes. An analogous issue was recently considered by the IMO’s Marine Environment Protection Committee (MEPC) with respect to proposed amendments to introduce certain exemptions from MARPOL Annexes I, IV, and VI survey and certification require117 Porathe/Prison/Man (note 22); Are E. Ottesen, Situation Awareness in Remote Operation of Autonomous Ships: Shore Control Center Guidelines, available at: https://www.ntnu.no/documents/ 10401/1264435841/Artikkel+Are+E+Ottesen.pdf/abb533ae-e73a-489e-80ec-f0e198e72c0a (accessed on 18 September 2017). 118

Annex I, Reg. 11 MARPOL.

AUTOMATED COMMERCIAL VESSELS

137

ments for unmanned non-self-propelled barges.119 The reasoning was that these unmanned barges do not have structures or equipment that generate pollutants addressed by MARPOL, such as sewage, and that the added administrative requirements, such as issuing and carrying the required International Sewage Pollution Prevention Certificate, were unnecessary. Among other, this was relevant for the implementation of the III Code in the conduct of the audits of Annexes for compliance with instruments covered by that code and port State inspection.120 The MEPC decided against an open-ended exemption, and favoured a capped exemption of no more than five years, despite views to the contrary.121 In addition to MARPOL, other environmental conventions could be impacted with respect to functions to be performed by the crew. In particular, the International Convention for the Control and Management of Ships’ Ballast Water and Sediments of 2004 prescribes that some officers and crew be familiar with the Ballast Water Management Plan and their duties in ballast water management operations on their ship.122 In the case of MASS this knowledge expectation will need to be transferred to the shore-based personnel or even performed autonomously.

V. Discussion This article has identified several issues concerning potential impacts of MASS on the international law of the sea and international maritime law. The pertinent UNCLOS jurisdictional frameworks relevant for MASS were set out, while the IMO conventions that nourish that framework and the nitty-gritty international rules and standards were considered. Functioning together, the two systems ensure that international navigation rights enjoyed by all ships are protected and that the exercise of jurisdiction over international shipping is within that framework and in accordance with generally accepted 119 Marine Environment Protection Committee (MEPC), Report of the Marine Environment Protection Committee on its 69th Session, IMO Doc. MEPC 69/21 (2016), 52. 120 IMO Assembly, IMO Instruments Implementation Code (III Code), Resolution A.1070(28) of 4 December 2013, IMO Doc. A 28/Res.1070 (2013). 121 MEPC (note 119), 52. Contrary views were expressed by China, Japan, Norway, and South Korea. Proposed amendments to MARPOL Annexes I, IV and VI to facilitate the exemption of survey and certification requirements for Unmanned Non-Self-Propelled (UNSP) barges, Submitted by Japan and the Republic of Korea, IMO Doc. MEPC 69/13/2 (2016). 122

Annex, Reg. B-6 BWM Convention.

138

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

international maritime rules and standards. The advent of MASS, although potentially accommodated, has been demonstrated to challenge the application of particular rules of the two systems. While there is room for interpreting affected rules in the conventions in a manner consistent with the general rules for the interpretation of treaties,123 there are also gaps that are not easily addressed through interpretation alone. Moreover, in the interest of clarity, amendment of a rule might be preferable to waiting for an interpretation from an authoritative body. The respective abilities of the UNCLOS and IMO conventions to respond to the change in the operational environment of shipping and navigation promised by MASS are different. In the event the manning requirements in Article 94 UNCLOS cannot be waived through a functional interpretation, it could be difficult to amend the Convention. While the UNCLOS has amendment procedures, they have not been utilised to date and change to the Convention has been effected through separate implementation agreements adopted after laborious diplomatic processes.124 However, the UNCLOS has a simplified procedure, as yet untested, that could potentially be used to address technical changes to Article 94. A State party would need to propose an amendment for adoption through the simplified procedure, communicated to the UN Secretary-General who in turn communicates the proposal to all States parties, and in the event of no objection being lodged by a State party within a period of one year, the amendment would be adopted.125 Should interpretation or amendment of Article 94 UNCLOS not be possible, future State practice in registering MASS vessels and exercising effective jurisdiction and control over them to the same extent as manned ships could well provide a customary law basis for accommodating automated ships.126 In comparison, IMO regulation is substantially more adaptable because several of the key instruments affected by MASS have tacit amendment procedures. The MEPC and MSC, as the two key committees tasked with the maintenance of the safety and environmental conventions for which the IMO is responsible, have the necessary authority and machinery to amend the technical aspects of safety and environmental conventions using 123

Art. 31 Vienna Convention on the Law of Treaties, 23 May 1969, UNTS 1155, 331.

124

Specifically with respect to Part VII’s provisions on straddling stocks and highly migratory species and Part XI provisions on deep seabed mining. 125 126

Art. 313 UNCLOS.

It is arguable that MASS, as a matter unregulated by the UNCLOS, may be governed by rules of general international law, see the UNCLOS preamble.

AUTOMATED COMMERCIAL VESSELS

139

the simplified tacit acceptance procedure following consideration by the pertinent committee, instead of a diplomatic conference convened by the Organization.127 However, a potential issue for the IMO is how to eliminate the manning requirement for fully autonomous ships in a manner that retains consistency with a provision to the contrary in Article 94 UNCLOS,128 unless the text can be interpreted to accommodate MASS as discussed earlier. The IMO is well poised and advised to proactively address the regulatory needs of MASS. As other studies, this article has observed that while much of the IMO regulatory system will be able to accommodate this new class(es) of vessels, it is very likely that there are many rules and standards in several maritime conventions that will need to be adapted through interpretation or amendment. The IMO’s strategic directions provide for keeping “under review the technical and operational safety aspects of all types of ships”,129 which will include MASS. It oversees the maintenance of a regulatory system to promote harmonisation and facilitate compliance with its instruments. Hence, the multi-Member State proposal for a regulatory scoping exercise of the Organization’s instruments is timely130 and the initial approach defined by the MSC appropriate.131 The IMO will take a proactive approach that will explore a range of impacts of surface vessels employing different levels of automation, including on safety, security, ports, pilotage, and incident response, including legal responsibility and liability issues.132 One can speculate on what precise regulatory changes will be necessary. Degrees of automation with respect to aspects of ship operation have been gradually introduced over time, but automation on a large scale as in the case of MASS has not yet been 127

For example: Art. 16 MARPOL; Art. VIII SOLAS Convention.

128

The IMO Committees “when exercising the functions conferred upon it by or under any international convention or other instrument, shall conform to the relevant provisions of the convention or instrument in question.”, Art. 31 (on the MSC) and Art. 41 (on the MEPC) Convention on the International Maritime Organization, 6 March 1948, UNTS 289, 3. The IMO policy has been that the adoption of an instrument is without prejudice to “the codification and development of the law of the sea in UNCLOS or any present or future claims and legal views of any State concerning the law of the sea and the nature and extent of coastal and flag State jurisdiction”, IMO Secretariat, Implications of the United Nations Convention on the Law of the Sea for the International Maritime Organization: Study by the Secretariat of the IMO, IMO Doc. LEG/MISC.7 (2012), 7. 129

IMO Assembly, Strategic Plan for the Organization (2016–2021), Resolution A.1097(29) of 1 December 2015, High-level Action 5.2.1. 130

MASS Scoping Exercise (note 18).

131

MSC (note 17), 79.

132

Ibid.

140

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

considered by the IMO. Given the multiple points of impact of MASS on maritime regulations discussed in this article, at a minimum the approach to adaptation should be systemic, rather than a piecemeal convention by convention, code by code approach. Hence, a comprehensive impact review of all of the IMO instruments is highly desirable and some studies have already made the first efforts in this direction.133 Changes to rules and standards concerning ship construction and equipping will be expected, in particular to take account of modifications to the usual designs for the bridge, equipment on board, crew accommodations and passageways, and life-saving craft. What is uncertain at this early stage is the extent of the change as we can reasonably expect different levels of automation on MASS, and across several classes of ships (e.g. bulkers, container ships). In this respect, the goal-based approach to maritime regulation will substantially facilitate the accommodation of various technologies to produce desired safety, security, and environmental outcomes.134 Particular MASS vessels can be expected to justify exemptions from certain rules. Some conventions have provisions permitting exemptions issued by national maritime administrations, and other States parties are expected to be notified of exemptions135 or equivalents.136 This flexibility will be useful in transitioning MASS into a system designed for very different ships. For example, compliance with particular certification requirements will be superfluous on ships with the highest levels of automation (e.g. sewage certificates), and exemptions would be appropriate. In other instances, rule compliance for MASS might need adaptation of the regulatory requirement. A fully automated vessel will likely have an electronic log. Consistently with the current trend, documentation

133

For example the CMI Working Paper (note 25).

134

IMO Assembly, Principles to be Considered when Drafting IMO Instruments, Resolution A.1103(29) of 26 November 2015, Annex. 135 For example Ch. I, Reg. 4(b) SOLAS Convention: “The Administration may exempt any ship which embodies features of a novel kind from any of the provisions of Chapters II-1, II-2, III and IV of these Regulations the application of which might seriously impede research into the development of such features and their incorporation in ships engaged on international voyages. Any such ship shall, however, comply with safety requirements which, in the opinion of that Administration, are adequate for the service for which it is intended and are such as to ensure the overall safety of the ship and which are acceptable to the Governments of the States to be visited by the ship. The Administration which allows any such exemption shall communicate to the Organization particulars of same and the reasons therefor which the Organization shall circulate to the Contracting Governments for their information”. 136

Ch. I, Reg. 5 SOLAS Convention.

AUTOMATED COMMERCIAL VESSELS

141

requirements will be satisfied electronically and reporting for various regulatory purposes will become increasingly passive. Classification societies will play a key role in helping to develop appropriate standards and unified requirements. IMO regulation frequently relies on the unified requirements adopted by the International Association of Classification Societies. However, regulatory accommodations will likely impact the mandatory audits under the III Code whose purpose is to promote compliance with international instruments with no more favourable treatment. The implications will extend to port State control inspections, which enforce international instruments also with no more favourable treatment to ship or flag. There will be implications for private law purposes which, while beyond the remit of this article, should be mentioned, such as the evolving meaning of seaworthiness in the carriage of goods by sea and insurance contracts, among others, since the ability of MASS vessels to depart on and complete the maritime adventure in a timely and safe manner will also depend on shore-based systems.

VI. Conclusion The advent of MASS is a potential game-changer for international shipping and its regulation. As the human factor is withdrawn from shipping in favour of artificial intelligence and accompanying automated systems, international maritime law is not likely to function on a business-as-usual scenario and will need to be adapted and further developed. However, it is unlikely that the human factor will be fully removed from shipping as there will continue to be trades and regions that will rely on manned ships. What can be envisioned is a future that will have a versatile mix of manned and unmanned ships, often interacting when in close proximity while navigating, varying by trading region, and with some trading regions having a higher concentration of automated ship operations because of early regional buy-in. MASS will potentially produce a range of direct and indirect impacts on the law of the sea and maritime law, in particular during their early introduction when the two legal regimes will need to adapt to the new technological environment. MASS will challenge a regulatory system that has often lagged behind technology. The IMO has an opportunity to put an anticipatory regulatory framework in place to influence the directions of the MASS technologies.

142

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

Finally, in scoping the regulatory impacts and future agenda, the IMO should also consider larger social questions. For example, as automated systems produce greater efficiencies and cost-savings, are the technological risks fully internalised? Is substitution of artificial intelligence for human judgement socially acceptable in all areas and aspects of shipping and navigation, for example also with respect to passenger vessels? Are safety valves against technological failure and hacking needed? How will MASS affect the distribution of risk in shipping and with what consequences for responsibility and liability?

International Civil Aviation and the Dehumanisation of Activities STEPHAN HOBE( AND BENJAMYN I. SCOTT((

ABSTRACT: International passenger and cargo transportation plays a crucial role in the world’s economy, and central to the success of this industry and to its continuous growth is ensuring safe, secure, and efficient operations at a national, regional, and international level. In pursuance of this, the aviation industry is becoming increasingly reliant on technology, with the aim of replacing the human user with technology. The introduction of automation, from its rudimentary beginnings, throughout its constant evolution, is now essential in almost every area of international aviation and is one of the core components in ensuring safe and efficient operations. However, its evolution, which has undoubtedly revolutionised the industry, has not always been free from challenges and controversies. This paper will therefore analyse, from both a manned and unmanned perspective, the benefits and challenges that automation has presented and continues to present to international civil aviation. KEYWORDS: Automatic, Autonomous, Aviation, Cybersecurity, Manned, Safety, Security, Unmanned

I. Introduction International passenger and cargo transportation plays a crucial role in the world’s economy with 0.9% of world gross domestic product predicted to be spent by consumers in 2017 on air transport, which amounts to 3,959 million passenger departures

(

LL.M., Dr. iur., Dr. habil., Professor for Public International Law, European Law, European and International Economic Law, and Director of the Institute of Air and Space Law, University of Cologne. ((

LL.B. (hon.), LL.M., LL.M. (adv.), Research Associate and Doctoral Candidate at the Institute of Air and Space Law, University of Cologne.

144

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

and 55.7 million freight tonnes.1 Central to the success of this industry and to its continuous growth is ensuring safe, secure, and efficient operations at a national, regional, and international level. In pursuance of this, aviation is becoming increasingly reliant on automated technology. The introduction of automation, from its rudimentary beginnings, throughout its constant evolution, is now essential in almost every area of international aviation and is one of the core components in ensuring safe, secure, and efficient operations. However, the evolution of automation, which has undoubtedly revolutionised the industry, has not always been free from challenges and controversies. This paper will therefore analyse, from both a manned and unmanned perspective, the benefits and challenges that automation has presented and continues to present to international aviation. Underlying this will be a threefold focus on the legal regime that has developed to regulate automated aviation technology and its use. First, the implementation and use of new technology must conform to the accepted level of safety. As a result, aviation safety aspects, such as flight crew training and airworthiness requirements, will have to be discussed. Second, such technology should be designed and used securely. Therefore, the relevant aviation security rules will be highlighted and critiqued, with specific attention being given to cybersecurity. Third, despite an extensive aviation safety and security legal regime, accidents and incidents may still occur, and because of this, the law must be capable of governing responsibility accordingly. Consequently, the paper will also examine how aviation rules have assigned responsibility in view of the varying degrees of automation. This three-fold focus will then allow for a holistic legal analysis of automated technology in aviation, which will be concluded with a case study on unmanned aircraft as this provides a real-time example of regulators trying to construct new rules for a fast developing, in some cases automatic or autonomous, aircraft sector.

1 International Air Transport Association (IATA), Economic Performance of the Airline Industry, 8 December 2016, available at: http://www.iata.org/whatwedo/Documents/economics/IATA-EconomicPerformance-of-the-Industry-end-year-2016-report.pdf (accessed on 18 April 2017).

INTERNATIONAL CIVIL AVIATION

145

II. Manned Aviation A. Terminology

No definition exists in the body of international air law, public or private, for ‘automation’ or other related terms. For example, the Chicago Convention on Internaional Civil Aviation of 1944 (Chicago Convention),2 with 192 signatory States (as of November 2017), which contains the overarching and underpinning legal framework that regulates international civil aviation, makes no reference to this concept. It is not surprising that the Chicago Convention does not contain such a definition in Article 96 on definitions and, instead, leaves it to the International Civil Aviation Organization (ICAO)3 to develop necessary definitions through Standards and Recommended Practices (SARPs),4 as expressed in the nineteen Annexes to the Chicago Convention.5 As opposed to the Convention, the Annexes contain numerous references to automation. By way of an example, Annex 2 on the Rules of the Air explicitly covers automation with the inclusion of ‘Automatic Dependent Surveillance-Broadcast (ADS-B)’: “A means by which aircraft, aerodrome vehicles and other objects can auto-

2

Convention on International Civil Aviation, 7 December 1944, UNTS 15, 295 (Chicago Convention).

3

“The International Civil Aviation Organization (ICAO) is a UN specialized agency, established by States in 1944 to manage the administration and governance of the Convention on International Civil Aviation (Chicago Convention).”, ICAO, About ICAO, available at: http://www.icao.int/abouticao/Pages/default.aspx (accessed on 18 April 2017). 4

States are required, as per Art. 37 Chicago Convention, to adopt regulations that are uniformed with those established under the Convention and Standards and Recommended Practices (SARPs) as found in the Annexes. ICAO has defined SARPs as follows: “a) Standard – any specification for physical characteristics, configuration, material, performance, personnel or procedure, the uniform application of which is recognized as necessary for the safety or regularity of international air navigation and to which Contracting States will conform in accordance with the Convention; in the event of impossibility of compliance, notification to the Council is compulsory under Article 38 of the Convention; and b) Recommended Practice – any specification for physical characteristics, configuration, material, performance, personnel or procedure, the uniform application of which is recognized as desirable in the interest of safety, regularity or efficiency of international air navigation and to which Contracting States will endeavour to conform in accordance with the Convention.”, ICAO, Resolution A36-13: Consolidated Statement of ICAO policies and associated practices related specifically to air navigation, Appendix A, in: ICAO, Assembly Resolutions in Force, Doc. 9902 (2007), II-2, II-3. 5 ICAO, The Convention on International Civil Aviation: Annexes 1 to 18, available at: http:// www.icao.int/safety/airnavigation/NationalityMarks/annexes_booklet_en.pdf (accessed on 18 April 2017). ICAO has yet to update the list to include Annex 19 on Safety Management.

146

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

matically transmit and/or receive data such as identification, position and additional data, as appropriate, in a broadcast mode via a data link”. Further, there is also indirect coverage of automation in Annex 2, such as inter alia, the inclusion of ‘Ground Visibility’: “The visibility at an aerodrome as reported by an accredited observer or by automatic systems.” Although there are direct and indirect references to automation, again, there is no attempt within the Annexes to specifically define the term. The usefulness of such a definition would, however, be questionable as, firstly, the legal status of the SARPs is highly debatable,6 so it is dubious whether such an inclusion would provide greater legal certainty. Secondly, it is doubtful whether such an inclusion is needed. This is because aviation regulation, which is often reactive in nature, would likely have produced a definition if one was required. Therefore, it is unlikely that a definition will be introduced in the near future due to a seeming lack of necessity. The lack of need for such a definition is also advanced by the implementation and interpretation of SARPs. This is because SARPs are designed to be implemented by States at a national level as prescribed in Articles 37 and 38 Chicago Convention.7 Therefore, the concept is open to interpretation by the States parties to the Convention as it is for the State to implement the SARPs, so a level of flexibility can be expected within the context of automation. In order to encompass the flexibility provided by ICAO and its usage in international aviation, a broad understanding of ‘automation’ will be adopted for the purpose of this essay. Therefore, ‘automation’ can be understood, for the purpose of this paper, as where a technology replaces, in whole or in part, or assists a human user.8 The replacement or assistance can equate to all levels, so it could include performance above that of the human or simply provide assistance that mitigates the burden placed on that user. 6 Pablo Mendes de Leon, An Introduction to Air Law (2012), 256. SARPs have a “conditional binding force” as, pursuant to Art. 38 Chicago Convention, “a contracting party must notify ICAO any time it fails to comply with a Standard in all respects; fails to bring its regulations or practices into full accord with any Standard; or adopts regulations or practices differing in any particular respect from the Standard.”, ICAO, Manual on Notification and Publication of Differences, Doc. 10055 (2015). 7 8

ICAO (note 6), para. 1.2.3.

Amy R. Pritchett, Aviation Automation: General Perspectives and Specific Guidance for the Design of Modes and Alerts, Reviews of Human Factors and Ergonomics 5 (1) (2009), 82.

INTERNATIONAL CIVIL AVIATION

147

B. Automation in Aviation

Automation in aviation should not be considered as one act, but rather as a multifaceted concept that can take many forms and be implemented in numerous ways. Accordingly, Raja Parasuraman and Thomas Sheridan identified four functions that automation can take which are identified below and then put into an aviation context. 1. Information acquisition: This is the process of obtaining data.9 For example, aircraft are fitted with instruments that measure certain factors, such as altitude and air speed. 2. Information analysis: This is information acquisition and then computing it into information which can be utilised by the human user. This can be, for example, predictive information, such as a fuel gauge which states how much fuel will be left in the tank at the place of arrival. This can then assist the ground-handlers responsible for re-fuelling.10 3. Decision and action selection: This is where the technology reviews the situation and provides recommendations. For example, the Traffic Alert and Collision Avoidance System (TCAS) will speak with another aircraft’s TCAS and then propose a cause of action to the pilots in order to avoid a collision. 4. Action implementation: This is where automation takes control based on received information. For example, when an aircraft senses wind shear,11 it will implement a go around manoeuvre to avoid the hazard.

9 Data refers to facts or figures, which once processed, interpreted, organised, structured, or presented becomes information, thus providing context to the data. 10

For the importance of fuel calculation, see the Gimli Glider Incident which involved Air Canada Flight 143 running out of fuel at 41,000 ft forcing it to conduct a gliding land. 11

“Wind shear is a change in wind speed and/or direction over a short distance. It can occur either horizontally or vertically and is most often associated with strong temperature inversions or density gradients. Wind shear can occur at high or low altitude.”, United States Federal Aviation Administration, Wind Shear, Doc. FAA-P-8740-40 AFS-8 (2008).

148

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

Furthermore, automation is an interaction between the human user and technology, and not every relationship will be the same. For example, as highlighted by Parasuraman and Sheridan,12 the levels, which can equally apply to aviation, are: 1. The automation technology offers no assistance and the user performs all the tasks; 2. the automation technology suggests alternative performance methods and the user must decide to accept or ignore them; or 3. the automation technology selects a way to perform the task, and; a) executes only after user approval, b) allows the user a set time to veto it before execution, c) executes it automatically and then informs the user, d) executes the suggestion automatically and then informs the user only if asked, or e) executes it and then ignores the user.13 It can be seen that automation, especially in such a technologically advanced and billion-dollar industry as aviation, is not a simple matter. Therefore, automation should be seen as an overarching term that encompasses numerous technologies, in multiple mediums, that carries out different roles and functions, and involves a complex relationship between the human user and technology.

C. Growth in Automation

1. Examples of Automation Automation in aviation is prevalent and a component in all of the main activities of aeronautical stakeholders. These include airlines, airports, air traffic control (ATC) 12 Raja Parasuraman/Thomas B. Sheridan, A Model for Types and Levels of Human Interaction with Automation, IEEE Transactions on Systems, Man, and Cybernetics – Part A: Systems and Human, 30 (3) (2000), 286. 13

Antonio Chialastri, Automation in Aviation, in: Florian Kongoli (ed.), Automation (2012), ch. 5.

INTERNATIONAL CIVIL AVIATION

149

and air space management entities, ground handling companies, and manufacturers. It is, therefore, not possible to comment on all areas of automation and how they are used by the different stakeholders. However, some of the more well-known tools used in aviation, which provide some context, are: 1. Auto-Pilot: This is a system used to control the trajectory of an aircraft without constant ‘hands-on’ control from the pilot. Auto-pilots do not replace a human operator, rather they assist in controlling the aircraft, allowing the pilot to focus on other elements of the flight. 2. Auto-Landing: This is a fully automated system that performs the landing procedure, in which the flight crew takes a supervisory role. This system allows an aircraft, for example, to land in bad weather conditions, such as fog and heavy rain, which would otherwise be difficult or impossible for a pilot to operate. 3. TCAS: This on-board system detects other aircraft in the area that could pose a collision hazard and then selects a mutually agreeable avoidance manoeuvre for the concerned aircraft. The flight crews must follow the instructions so as to avoid the collision. 4. Enhanced Ground Proximity Warning System: This is a similar system to TCAS, which is designed to prevent collisions with the terrain.14 5. Airport Baggage Systems: Airports often have fully integrated automated high-speed airport baggage handling systems. These systems include security screening, storage, allocation, and transportation of baggage (e.g. from check-in to the departure gate, transfers, and aircraft to arrival haul). It can be seen from the above examples that automation technology is not always for the use of one stakeholder, such as TCAS, which has a useful application for the airlines, the people on-board of the aircraft, other airspace users, and ATC. Furthermore, automation technology can be used for a variety of outcomes, such as safety, minimising workload, commercial benefits, and security. Thus, each technology must be assessed on its own merits with regard to its level of automation, its function, and its intended user, which adds to the multifaceted application of automation in aviation.

14

Pritchett (note 8).

150

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

2. The Evolution of Automation In aviation, where one mistake can lead to the total loss of an aircraft, it is preferable for critical reliable technology to be introduced at a prudent pace, rather than it being state-of-the-art but still posing potential deficiencies due to a lack of testing in an operational environment. As a result, advancements in aviation technology frequently follow a slow evolutionary path, as opposed to a fast revolutionary one. Automation seems to follow an evolutionary path rather than a revolutionary approach. Its adoption on board aircrafts does not respond to the planned purpose of enhancing safety “from scratch” in a consistent way, but rather resembles a biological organism trying to continuously adapt to the challenges posed by its environment (fly-fix-fly).15

This trial-and-response approach to automation has produced three main generations of automation within aviation: mechanical, electrical, and electronic. At the advent of civil aviation, there were very few instruments on-board an aircraft for the pilot to utilise. For example, navigation had to be done via visual line of sight (VLOS), maps, and notable landmarks. With the advancement of technology, pushed by humankind’s realisation of the usefulness of aircraft, mechanical instruments were introduced into the cockpit. Such initial instalments included anemometers and altimeters, which provided basic information on orientation and environmental conditions. Rudimentary automation was thus introduced on-board the aircraft “from 1920 to 1930, in the form of an autopilot based on a mechanical engineering concept that was designed to keep the aircraft flying straight: a very basic input to control the flight at a ‘skill’ level.”16 As aircraft became larger and more complex, the need for mechanical support became increasingly necessary. For example, “[i]nstead of direct control, with the yoke mechanically attached to the ailerons, airplanes began to be constructed with a series of mechanisms intervening between the pilot’s input and the expected output.”17 This initial stage of mechanical automation began the first steps of widening the “gap

15

Chialastri (note 13), 84.

16

Ibid., 85.

17

Ibid.

INTERNATIONAL CIVIL AVIATION

151

between the pilot’s input (action on the yoke) and the final outcome (aerodynamic movement).”18 As electricity became more advanced, the cockpit saw a replacement of mechanical devices with electrical ones. This marked the beginning of the second generation where electrical instruments were installed to increase automation. This took-off in the 1960s, which included: electric autopilots, auto-throttle (to manage the power setting in order to maintain a selected speed, or a vertical speed), flight directors (used to show pilots how to manoeuvre to achieve a pre-selected target such as speed, path-tracking and so forth), airborne weather radars, navigation instruments, inertial platforms, but also improved alarming and warning systems capable of detecting several parameters of engines and other equipment.19

As well as assisting the pilot’s skill-based level, as with the first generation, the second generation also assisted with compliance with the law. This level “indicates compliance with the rules, norms, laws, and everything laid down in the official documentation.”20 For example, to ensure compliance with fixed regulatory parameters, instruments were fitted with alarms to indicate to the pilot when these were close to being met. The third generation was predominantly driven by the electronic revolution occurring from the mid-80s. Electronics significantly helped to diminish the clutter of instruments on board and allowed for replacing old indicators – gauges in the form of round-dial, black and white mechanical indicators for every monitored parameter – with integrated coloured […] capable of providing a synthetic and analytic view of multiple parameters in a limited area of the cockpit.21

The introduction of electronic visual displays meant that not every instrument or piece of information needed to be on display for the user. This uncluttered the working environment, but meant that not every piece of information is automatically accessible. Thus, this necessitated conscious checking and coordination, resulting in

18

Ibid.

19

Ibid.

20

Ibid., 83.

21

Ibid., 86.

152

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

the user increasingly being expected to know which button to press and what the probable outcome is likely to be, rather than to actually understand the process. The evolution of automation is not over and examples from other industries may highlight how it will advance in aviation. For example, in the automotive industry, Waymo, a company under Google’s parent company Alphabet Inc., is developing an autonomous car. The car is being designed to be ‘self-drive’ which requires very little input from the passengers and rather utilises technology to execute the drive aspect. As seen with the automotive industry, full automation may also be the next step in the evolution of aviation. Steps can already be seen within the aviation industry in line with automotive automation. For example, Uber, the on-demand taxi company, has proposed a pilotless vertical take-off and landing aircraft taxi that will only require the passenger to input the destination before the aircraft takes over.22

3. Why is Automation Important? Within aviation, two primary drivers are behind automation. The first one is the increased level of safety that it can provide. Safety, as identified throughout this paper, is of key importance to the aviation industry, of which automation is a significant component in its enhancement. For example, systems monitoring the condition of the aircraft allow pilots and maintenance staff to know the health of the aircraft without having to wait for a scheduled safety inspection to take place. Therefore, faults can be identified as soon as they occur, which allows them to be fixed before they pose a serious risk to safety. Another example, within the scope of navigation, is that automation has also relieved flight crews and the ATC of repetitive or unrewarding tasks, and of actions where human performance is less effective, so they can focus on more critical tasks. The second fundamental driver for automation are economic factors. A crucial item in an airline’s balance sheet is the cost of fuel. Saving on fuel is vital to remain competitive in the market. According to the International Air Transports Association’s (IATA) estimates, “a one percent reduction in fuel consumption translates into annual savings amounting to 100,000,000 dollars a year for IATA carriers of a partic22

Uber, Fast-Forwarding to a Future of On-Demand Urban Air Transportation, 27 October 2016, available at: https://www.uber.com/elevate.pdf (accessed on 18 April 2017).

INTERNATIONAL CIVIL AVIATION

153

ular State”.23 A large contributor to increased fuel efficiency is the reduction in weight caused by the removal of heavy steel cables and machines, for lighter computers and wires that came with the second and third generation of automation. A further example is that aircraft are fitted with instruments that measure fuel consumption during a flight which can then be reviewed to see where the pilot can improve on fuel efficiency and thus reduce costs. Maintenance costs are also significantly reduced through automation. For example, a single computer can now fulfill the function of several machines. This significantly reduces the number of different parts in an aircraft that may need to be repaired or replaced in case of a fault. Therefore, repairing or replacing a single computer, rather than potentially every individualised component, means that fewer spare parts are required to be at hand which takes up storage space and costs money to purchase. Furthermore, training of maintenance personnel is simplified as it can now focus on multipurpose technology, rather than on a broad range of specialised items. As many functions are taken over by computers, there is less demand on the knowledge base of the pilot. It then takes less time and fewer resources to train pilots. This then cuts costs on initial recruitment and continuing training. Furthermore, cockpits are becoming more standardised so it is easier for pilots to transition between aircraft types. Therefore, pilots are more flexible and can move more easily between the different aircraft types that an airline may have, rather than requiring specific and fixed pilots per type. This could then lower the number of employees. Automation has also allowed for greater operational flexibility. For example, it would be difficult to land an aircraft in bad weather conditions, such as heavy fog. However, auto-land and other tools can assist the pilot in this activity. Therefore, this reduces delays and cancellations, which can have financial consequences,24 and can allow aircraft to fly more often allowing the airline to sell more tickets. Automation has

23 24

ICAO, Human Factors Training Manual, Doc. 9683-AN/950 (1998), para. 3.1.3.

EC Regulation 261/2004 of the European Parliament and of the Council of 11 February 2004 establishing common rules on compensation and assistance to passengers in the event of denied boarding and of cancellation or long delay of flights, OJ L 46, 1, and repealing EEC Regulation 295/91 of 8 February 1991, OJ L 36, 5.

154

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

also streamlined aircraft turn around at airports.25 For example, technology facilitates low cost airlines like Ryanair to turn around within 30 minutes.

D. Issues

Since the inception of aviation and its increased automation, several issues have persisted to varying degrees which manufacturers, regulators, and users have tried to remedy or at least mitigate. First, aviation involves highly complicated technologies that require significant initial and continuous training by those involved, so that the technology can be safely operated. To ensure this, laws have been adopted. For example, within the European Union (EU), Article 7 and Annex III of Regulation 216/2008 provide for the adoption of implementing rules necessary “for establishing the conditions for certifying pilots as well as persons involved in their training, testing or checking, for the attestation of cabin crew members and for the assessment of their medical fitness.” Consequently, Commission Regulation 1178/201126 provides technical rules for pilots and cabin crew, and those involved in their training and medical assessments. A significant amount of training is, thus, required to demonstrate that the users can safely use the technology, which takes time and costs money. Second, the automation of aviation has been an evolutionary process and, with this, those who use the technology have also had to evolve. Therefore, the user has to adapt to automotive technology, from time to time, rather than learning in a complete system. This can have adjustment issues as it takes the user time to become familiar and comfortable with the new technology, and for its use to become instinctive, as opposed to the older technology. Third, the more tasks assigned to technology and taken away from the user, the less the user begins to understand or needs to understand that technology. This can create 25 ‘Turnaround’ refers to the time needed to unload passengers, baggage, and cargo from the previous flight and to then load the new passengers, baggage, and cargo for the next flight, as well as any refuelling and servicing. 26

Europan Union (EU) Commission Regulation 1178/2011 of 3 November 2011, OJ L 311, 1, laying down technical requirements and administrative procedures related to civil aviation aircrew pursuant to EC Regulation 216/2008 of 19 March 2008, OJ L 79, 1 of the European Parliament and of the Council (amended).

INTERNATIONAL CIVIL AVIATION

155

complacency where the user simply allows the technology to do its job without questioning or understanding it. Further, the basic skills of the user may decline as a result of lack of practice. This could then produce safety risks as the user is no longer able to perform the assigned task without the input from automated technology, which could prove devastating if the technology fails. Fourth, there is still a need for users. Automation is not always designed to function in every possible situation; rather, boundary conditions are prescribed which in turn limit the competencies of the technology. Therefore, automation may be appropriate under normal conditions; however, these are not always present in aviation and a user may be required to step in to combat events outside of the technology’s competencies. The scope of the technology should then be appreciated, with the technology being appropriately applied and balanced against the need for a user. Fifth, it may not always be appropriate to automate a function. For example, it may be the case that a human can deal with a task in a safer and more appropriate way than a piece of technology. Additionally, in cases where a relationship between the technology and the user is required, the two would have to speak a single language in order to communicate. This may result in oversimplifications of the communicated information as a user may not be able to process all of the information available to the computer. Thus, it may be more appropriate to avoid automation in this area or, in turn, replace the human activity fully with automation. Sixth, although a piece of technology may be fully tested and deemed safe, following a scrutinising certification process, such as through Regulation 748/2012 on Initial Airworthiness and Regulation 1321/2014 on Continuing Airworthiness, its social acceptance may not yet be present. For example, while the cockpit is becoming more automated, it may be a step too far in the near future for a pilotless cockpit when passengers are involved. This may change in the future, as, for example, the Elevate, which is a pilotless rotorcraft taxi proposed by Uber, may pave the way for increased social acceptance. Finally, an initiated automated process may be overridden by the user or certain warnings may be ignored. For example, one function of automation is to warn of errors and failures. However, not all errors and failures produce the same risk to safety and should then be responded to appropriately. Also, the technology may be overly sensitive, so it notifies unnecessarily. Therefore, there is often a level of discretion left

156

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

to the user and this may be applied inconsistently. Technology is not always correct and this can result in the user looking for non-existing issues or ignoring a warning. This can result in the validity of the technology being seriously degraded. A significant example of some of the issues highlighted above is the Überlingen mid-air collision of 1 July 2002. The crash involved Bashkirian Airlines Flight 2937, a Tupolev Tu-154, which was a chartered flight from Moscow, Russia, to Barcelona, Spain, carrying 60 passengers (45 of the passengers were Russian schoolchildren) and nine crew members. The second aircraft was DHL Flight 611, which was a Boeing 757 cargo aircraft that had originated in Bahrain and was being flown from Bergamo, Italy, to Brussels, Belgium by two crew members. In this case, both aircraft where flying at 36,000 ft (10,973 m) over German territory. The airspace was, however, controlled by Skyguide, a Swiss company, from Zurich. Numerous factors contributed to the mid-air collision, one of them being the conflicting information received by the pilots. The ATC, after noticing that the two aircraft were on a collision course, instructed Flight 2937 to descend by 1000 ft. This command, however, contradicted the TCAS instruction which told the pilot to climb. Flight 611’s TCAS system instructed the pilot to descend. If both pilots had followed the TCAS instructions, then the collision would not have occurred. However, the instructions of the ATC, rather than the TCAS, were followed by the Russian aircraft and the aircraft remained on a collision course. The two aircraft collided in mid-air, killing all on-board. This accident highlights the importance of proper integration of automation technologies and the requirement for proper training of the users.28 This case also highlights the need for automation technology on-board aircraft, but also that its application should be harmonised and that the users should be adequately trained, otherwise its presence is ineffective. 27

27

Art. 2 Chicago Convention: “For the purposes of this Convention the territory of a State shall be deemed to be the land areas and territorial waters adjacent thereto under the sovereignty, suzerainty, protection or mandate of such State”. 28 German Federal Bureau of Aircraft Accident Investigation, Investigation Report, May 2004, available at: http://www.bfu-web.de/EN/Publications/Investigation%20Report/2002/Report_02_ AX001-1-2_Ueberlingen_Report.pdf?__blob=publicationFile (accessed on 18 April 2017).

INTERNATIONAL CIVIL AVIATION

157

E. Growing Issue – Legal Issues of Cybersecurity

The regulation of aviation is often a reactive exercise that follows an accident or incident, or the advancement of technology. It is, then, more often than not the role of the regulator to play catch-up with events or the industry. This has often been the case with automation technologies. For example, the tragic event over Überlingen prompted additional requirements for training. The creation of new technologies has also seen an expansion in the ICAO Annexes and the EU technical Regulations within the European Aviation Safety Agency (EASA)29 legal system centred on aviation safety and environmental protection. One of the most critical areas that the regulators are now facing, having both an impact on safety and security, is cyberattacks and its mitigation via cybersecurity.30 Within the aviation industry, systems must simultaneously and constantly speak with one another. For example, computers enable passengers to book tickets, check-in online and clear airport security. Furthermore, air carriers rely on them in order to operate their aircraft, such as, through varying communications links with ground handlers, air traffic controllers, other air users and the relevant airports.31

Furthermore, the issue of cybersecurity in aviation is not stationary as the “aviation industry is moving toward a more networked environment to improve all facets of air travel, from an aircraft on the ground to take off to air flight” and “the more complex and advanced a network becomes, the more technical glitches and vulnerabilities it 29 The European Aviation Safety Agency (EASA) is a specialised Agency of the EU. It was established in 2002 by EC Regulation 1592/2002 of 7 September 2002, OJ L 240, 1, which was then repealed by EC Regulation 216/2008 of 19 March 2008, OJ L 79, 1. Art. 2 (1) Regulation 216/2008 states that the main objective of the Agency is to “establish and maintain a high uniform level of civil aviation safety in Europe”. Art. 2 (2) then sets forth a number of additional objectives in the areas of certification, rulemaking, standardisation, and research and safety analysis, EC Regulation 216/2008 on common rules in the field of civil aviation and establishing a European Aviation Safety Agency, and repealing Council Directive 91/670/EEC of 31 December 1991, OJ L 373, 21, EC Regulation 1592/2002 of 7 September 2002, OJ L 240, 1 and Directive 2004/36/E of 30 April 2004, OJ L 143, 76. 30

Cybersecurity can be defined as “[t]he activity or process, ability or capability, or state whereby information and communications systems and the information contained therein are protected from and/or defended against damage, unauthorized use or modification, or exploitation.”, National Initiative for Cybersecurity Careers and Studies, Glossary, available at: https://www.niccs.us-cert.gov/glossary (accessed on 18 April 2017). 31

Benjamyn I. Scott/Andrea Trimarchi, The Digital Aviation Industry: A Balancing Act between Cybersecurity and European Consumer Protection, Air and Space Law 42 (2017), 443, 444.

158

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

contains, and the more difficult it becomes to manage from a security perspective.”32 Therfore, computers and machines which replace the user are becoming increasingly abundant within this industry and, as these must speak with one another, cybersecurity will continue to grow in importance as it will be vital in providing safe and secure transportation. While some attempts have been made to regulate these issues, the regulators at both international and regional levels are only beginning to comprehensively tackle this matter. At the international level, steps have already been taken to bring safety aspects into discussions on cybersecurity as, for example, on 26 February 2014 the ICAO Council adopted Amendment 12 to Annex 17 which included Recommended Practices relating to cybersecurity. 4.9.1 Recommendation – Each Contracting State should, in accordance with the risk assessment carried out by its relevant national authorities, ensure that measures are developed in order to protect critical information and communications technology systems used for civil aviation purposes from interference that may jeopardize the safety of civil aviation. 4.9.2 Recommendation – Each Contracting State should encourage entities involved with or responsible for the implementation of various aspects of the national civil aviation security programme to identify their critical information and communications technology systems, including threats and vulnerabilities thereto, and develop protective measures to include, inter alia, security by design, supply chain security, network separation, and remote access control, as appropriate.

Amendment 12 breaks away from the normal phraseology of Annex 17, which predominately refers to ‘unlawful interference’. If this phrasing was introduced into the Amendment, it would have resulted in a significant limitation to its scope as it only comprises illegal acts. This approach is adopted in the international conventions, drafted under the auspices of ICAO, on sabotage33 and offences on board aircraft,34 which 32 Emilio Iasiello, Aviation and Cyberspace: Convergence of Domains, Convergence of Threats, Air and Space Power Journal (2013), 3, 7. 33 Convention for the Suppression of Unlawful Acts against the Safety of Civil Aviation, 23 September 1971, UNTS 974, 177 (Montreal Convention 1971) and the Protocol for the Suppression of Unlawful Acts of Violence at Airports Serving International Civil Aviation, 23 September 1971, UNTS 1589, 474 (Montreal Protocol 1988). 34

Convention on Offences and Certain Other Acts Committed on Board Aircraft, 14 September 1963, UNTS 704, 216 (Tokyo Convention) and the Protocol to Amend the Convention on Offences and Certain Other Acts Committed on Board Aircraft, 4 April 2014 (not in force). Convention for the Suppression of Unlawful Seizure of Aircraft, 16 December 1970, UNTS 860, 106, and the Protocol Supplementary to the Convention for the Suppression of Unlawful Seizure of Aircraft, 10 September 2010 (not in force).

INTERNATIONAL CIVIL AVIATION

159

may have some relevance for cyberattacks. Rather, the Amendment takes a broader approach that focuses on the technology, as opposed to an unauthorised human user. This then directs attention towards securing the technology from outside interference, which will help ensure that the automated acts of the technology are those that are intended. The regulatory focus adopted by ICAO, which predominantly focuses on unlawful interference, was also originally adopted under the EU acquis through EC Regulation 2320/2002, as repealed by EC Regulation 300/2008 on common rules in the field of civil aviation security. However, the Draft Proposal to revise Regulation 216/200835 on aviation safety has allowed the EU to tighten the relationship between cybersecurity and aviation safety, which has a knock-on effect for automated technology. Within this Draft Proposal, Article 76 (1) would provide competencies to the Agency to cooperate with Member States and the Commission on security, including cybersecurity, matters related to civil aviation. Article 76 (2) then affirms and strengthens the connection between Regulation 300/2008 on aviation security and the EASA safety system as it gives the Agency competencies to provide technical assistance to the European Commission in the implementation of the security Regulation. Finally, Article 76 (3) allows the Agency to take certain measures to protect civil aviation against acts of unlawful interference. The Annexes to the Draft Proposal Regulation also contain changes that render cybersecurity part of traditional safety matters. The Draft Proposal Regulation would arguably bring cybersecurity under the whole EU aviation safety regulatory system. It is expected that amendments will also be needed in the Implementing Regulations and Annexes to Regulation 216/2008. The EU approach attempts to build upon the work already done by ICAO, by covering both technical requirements and assisting with the regulation of unlawful interference. Both are important components in ensuring that technology, including automated technology, functions as is intended. However, the Draft Proposal has yet to enter into force, so the current regulatory environment is lacking any comprehensive rules, and may be subject to revision during the legislative process. Thus, there is a lacuna in the law with regard to automated technology’s vulnerability to cyberattacks which is an indicative 35 European Commission, Proposal for a Regulation of the European Parliament and of the Council on common rules in the field of civil aviation and establishing a European Union Aviation Safety Agency, and repealing EC Regulation 216/2008, 19 March 2008, OJ L 79, 1, of the European Parliament and of the Council, COM(2015) 613 final 2015/0277 (COD), 7 December 2015. This has been affirmed by the European Council’s document of the same name, 2016/0277 (COD), 11 January 2018.

160

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

example of the slow-moving pace of aviation regulation and compromises the usefulness of such automated technology in enhancing aviation safety.

III. Unmanned Aircraft A. Introduction36

The advent of aviation is often fixed on 19 October 1783 when the Montgolfier brothers (Joseph-Michel and Jacques-Étienne) launched the first manned flight in Paris, France, via a tethered balloon or on 17 December 1903 when the Wright brothers (Orville and Wilbur) carried out the first sustained, controlled, powered, heavierthan-air manned flight in North Carolina, USA. However, it is not always acknowledged that unmanned aviation predates that of manned flight, with an early example including kites used in China as far back as the fifth century BC.37 While unmanned aircraft have been utilised for centuries, it was the twentieth century, with the advancement in aeronautical, automated, and communicatory technologies, that saw notable growth in their application.38 The history of unmanned aircraft is rooted in the military context which is still prevalent today, however, the 21st century has brought about a dramatic increase in non-military applications of unmanned aircraft.39 Unmanned aircraft are then becoming more relevant for civil use and this trend is expected to continue. Current unmanned aircraft already utilise automated technology. For example, large unmanned aircraft are equipped with sophisticated autopilots that allow them to traverse the globe to conduct remote activities. However, small commercially purchased unmanned aircraft may also be fitted with such equipment. This could include 36 Benjamyn I. Scott, Overview, in: id. (ed.), The Law of Unmanned Aircraft Systems: An Introduction to the Current and Future Regulation of UAS under National, Regional and International Law (2016), 3. 37

Konstantinos Dalamagkidis/Kimon P. Valavanis/Les A. Piegl, On Integrating Unmanned Aircraft Systems into the National Airspace System (2012). 38 39

For an overview of the history of unmanned aircraft see Scott (note 36).

Pablo Mendes de Leon/Benjamyn I. Scott, An Analysis of Unmanned Aircraft Systems under Air Law, in: Aleš Završnik (ed.), Drones and Unmanned Aerial Systems: Legal and Social Implications for Security and Surveillance (2016), 191.

INTERNATIONAL CIVIL AVIATION

161

basic detect and avoid systems, meteorology analysing tools that measure wind currents to ensure stability, and default land settings that initiate when the aircraft senses that it is no longer capable of safe flight. As technology advances and becomes more accessible, automation of unmanned aircraft is likely to increase. As a result of the growing relevance of unmanned aircraft within civil use, the European Commission has claimed that the unmanned aircraft sector, within the next twenty years, is predicted in the EU to “directly employ more than 100,000 people [and] have an economic impact exceeding EUR 10 billion per year, mainly in services.”40 Despite this, “[there] is, however, currently an absence of a clear legal framework governing unmanned aircraft across the world; therefore, steps are being taken by the national authorities, regional entities and international organisations to address this issue.”41 The regulation of this highly automated aviation technology is thus in a transition period that is surrounded by legal uncertainty and must therefore be examined in order to determine how regulators are handling unmanned automation and highlight any issues that automation is posing to the regulators.

B. Definitions

The most ubiquitous term used to refer to unmanned aircraft is the term ‘drone’. It is the term colloquially used by the media, industry, and the public, and also by regulators such as EASA to refer generally to these aircraft. However, within official legal use, a more technical and less politically sensitive term is often preferred. As a result of this, several terms have appeared in national legislation, regional regulatory steps, and legal commentaries. It is important to note that not every term pertains to the same content, which could have consequences for automation. There has been a significant terminological shift that has recently occurred which has seen the term ‘unmanned aircraft system’ (UAS) being used and defined. The EU, in its Notice of Proposed Amendment (NPA) on the ‘Introduction of a regulatory framework for the operation of drones’, a regulatory step taken by EASA towards the production of a Commission Regulation that will contain the specific technical rules 40 European Commission, Unmanned Aircraft, available at: http://ec.europa.eu/growth/sectors/ aeronautics/rpas_en (accessed on 18 April 2017). 41

Scott (note 36), 5.

162

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

on the ‘open’ and ‘specific’ categories of UAS,42 has defined this to mean “the UA and any equipment, apparatus, appurtenance, software or accessory that is necessary for the safe operation of the UA.”43 This term, therefore, encapsulates the importance of other components beyond just the aircraft; it is a whole system term. ‘Unmanned aircraft’ (UA), on the other hand, is often used within the term UAS to specifically refer to only the aircraft. The Draft Proposal for the revision of Regulation 216/2008 defines ‘unmanned aircraft’ to mean “any aircraft operated or designed to be operated without a pilot on board”.44 UAS is widely used by the EU,45 including EASA,46 ICAO,47 and by some national aviation authorities. The term UAS is broad and inclusive as it does not delimit its scope based on the type of control exercised by a pilot. This is, however, not the case with all terms, of which two subcategories of unmanned aircraft have emerged. One subcategory can be referred to as ‘remotely piloted aircraft systems’ (RPAS). This is a total systems term, like UAS, that includes the remotely piloted aircraft (RPA), the operator, communication and data links, satellites, ground control stations, additional staff, support systems and any other part required to operate the aircraft.48

In other words, an RPAS cannot be fully automatic and must, instead, be controlled by a human pilot from a remote location.49 The second subcategory, which has sometimes been labelled ‘unmanned drone’,50 refers to automated unmanned aircraft. This 42

See infra, III. D.

43

EASA, Introduction of a regulatory framework for the operation of drones: Unmanned aircraft system operations in the open and specific category, Notice of Proposed Amendment, NPA 2017-05 (A), Art. 2 (cc). NPA 2017-05 (A) replaced EASA, ‘Prototype’ Commission Regulation on Unmanned Aircraft Operations, 22 August 2016, Art. 2 (v) available at: https://www.easa.europa.eu/system/files/dfu/ UAS%20Prototype%20Regulation%20final.pdf (accessed on 18 April 2017). This has been affirmed, with minor changes, by EASA’s Opinion 1/2018, available at: https://www.easa.europa.eu/sites/default/ files/dfu/Opinion%20No%2001-2018.pdf (accessed on 27 February 2018). 44

European Commission (note 35).

45

Id., Unmanned Aircraft Systems (UAS), available at: http://ec.europa.eu/transport/modes/air/ uas/index_en.htm (accessed on 18 April 2017). 46 See EASA, Policy Statement Airworthiness Certification of Unmanned Aircraft Systems (UAS), Doc. E.Y01301 (2009), and ICAO, Unmanned Aircraft Systems (UAS), Cir 328 AN/190 (2011). 47

ICAO (note 46).

48

Benjamyn I. Scott, Terminology, Definitions and Classifications, in: id. (ed.) (note 36), 11.

49

ICAO, Manual on Remotely Piloted Aircraft Systems (RPAS), Draft v. 1.0, 2 April 2012, 22.

50

European Commission, Remotely Piloted Aviation Systems (RPAS): Frequently Asked Questions, Memo, 8 April 2014, 1.

INTERNATIONAL CIVIL AVIATION

163

may be defined as “automatically programmed – without being piloted, even remotely”.51 The removal of the remote pilot from the system has produced some confusion as the words ‘automatic’ and ‘autonomous’, which correspond to different levels of control, are often used interchangeably. An ‘automatic UAS operation’, as proposed by EASA in its NPA, “means an operation following preprogrammed instructions that the UAS executes while the remote pilot is able to intervene at any time”.52 This would include unmanned aircraft that would commence a flight following some data pre-input before beginning the flight. On the other hand, ‘autonomous unmanned aircraft’ may be defined, as proposed by EASA in its NPA, as an “unmanned aircraft that does not allow pilot intervention in the management of the flight”.53 They do not act in a predetermined way, as with automatic unmanned aircraft, and “rather they act in a pragmatic way by using the environment to make decisions.”54 This could include such unmanned aircraft that track and film people engaged in sports activities. These aircraft use tracking technology in order to stay close to the target and other technologies to monitor for hazards and to keep stabilised flight suitable for filming. The inclusion of automated technologies within unmanned aircraft has been noted, and this has resulted in different terms being used to express the varying levels. Therefore, caution must be given as not every term refers to an unmanned aircraft with automated technology. The EU has tried to remedy any ambiguity as the definitions presented in the EASA NPA differentiate between piloted, automatic, and autonomous. This was done so that the EU rules would be as broad as possible and thus encompassing most, if not all, unmanned aircraft. ICAO has also made it explicitly clear in its ‘Manual on Remotely Piloted Aircraft Systems (RPAS)’55 that “[a]ll unmanned aircraft, whether remotely piloted, fully autonomous or combinations thereof, are subject to the provisions of Article 8 of the Convention on International Civil Air51

Ibid.

52

EASA, NPA (note 43), Art. 2 (c).

53

ICAO (note 46), ix. EASA has also defined this to mean “an operation during which a UA operates without the possibility for remote-pilot intervention in the management of the flight”, EASA, NPA (note 43), Art. 2 (d). 54 55

Scott (note 48), 12.

ICAO, Manual on Remotely Piloted Aircraft Systems (RPAS), Doc. 10019 AN/507 (2015), paras. 1–8.

164

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

craft”.56 However, this article will demonstrate below that this inclusive attitude of the regulators has produced some specific problems for the regulation of automated unmanned aircraft.

C. Applicability to Current International Air Law

The Chicago Convention, as shown above, is relevant to international aviation and, as a result, may contain some provisions that affect automated unmanned aircraft. Before assessing its relevance in this context, we must note that the Chicago Convention has certain limitations and these must be assessed to determine its applicability. First, the Chicago Convention only applies to international air transport. As the majority of current civil unmanned aircraft activities are only domestic, this criterion is usually not satisfied. This will likely change as technology advances, regulations change, and as the use of unmanned aircraft evolves. Second, an aircraft, pursuant to ICAO Annexes, is “[a]ny machine that can derive support in the atmosphere from the reactions of the air other than the reactions of the air against the earth’s surface.”57 In order for the Chicago Convention to apply to unmanned aircraft, they must satisfy this definition which includes winged, rotor, and lighter-than-air unmanned vehicles.58 Third, the Chicago Convention is applicable to civil aircraft pursuant to Article 3. State aircraft are those used “in military, customs and police services” and similar services which can be demonstrated via State practice.59 Civil aircraft is then defined negatively, to mean that any which are not military are deened to be civil, which condition many unmanned aircraft satisfy. Therefore, some unmanned aircraft may fall under the scope of the Chicago Convention. The most explicit article contained in the Chicago Convention is Article 8 which is on ‘Pilotless Aircraft’:

56

Ibid., para. 1.1.

57

Id., Annex 2 to the Convention on International Civil Aviation: Rules of the Air (10th ed. July 2005), ch. 1 on Definitions, available at: https://www.icao.int/Meetings/anconf12/Document%20 Archive/an02_cons%5B1%5D.pdf (accessed on 21 September 2017). 58

See id. (note 49).

59

Art. 3 (b) Chicago Convention.

INTERNATIONAL CIVIL AVIATION

165

No aircraft capable of being flown without a pilot shall be flown without a pilot over the territory of a contracting State without special authorization by that State and in accordance with the terms of such authorization. Each contracting State undertakes to insure that the flight of such aircraft without a pilot in regions open to civil aircraft shall be so controlled as to obviate danger to civil aircraft.

The obligation that special authorisation must be obtained and that dangers to civil aviation must be controlled is clearly established. However, the scope of the Article is less apparent. This is because the term ‘pilotless aircraft’, which refers to aircraft “capable of being flown without a pilot”, is open to broad interpretation. ICAO has maintained the position that this term is to be understood to refer to RPAS and automated unmanned aircraft.60 However, the term ‘pilotless’ could indicate the absence of a pilot on board and outside of the aircraft which would then only include fully automated and automatic unmanned aircraft. While this point is easily overlooked as ICAO is continuing its activities with the full spectrum of unmanned aircraft, it at least shows that the existing aviation rules, which were written with manned aviation prominently in mind, are not necessarily best suited to deal with the specific complex situation of unmanned aircraft. This is because, for example, here there is a clear ambiguity with regard to the application of Article 8 as it is not overtly clear if it applies to piloted, automated, or autonomous unmanned aircraft.

D. Some Issues Encountered in Europe

Annex II of Regulation 216/2008 currently restricts EU competencies on regulating civil unmanned aircraft with a Maximum Take-Off Mass of #150 kg. As a result of this, the rules on unmanned aircraft in Europe are not harmonised and are instead left to national law to regulate. As this is having detrimental consequences to safety and commercial interests, with some EU Member States yet to enact national laws, there has been an awakening to remedy this regulatory patchwork. The Riga Declaration, which was endorsed by the Transport Commissioner and the Directorate-General for Mobility and Transport of the European Commission,61 stat60 See ICAO (note 46). See Mikko Huttunen, Unmanned, Remotely Piloted, or Something Else? Analysing the Terminological Dogfight, Air and Space Law 42 (3) (2017), 349. 61 Riga Declaration on Remotely Piloted Aircraft (Drones), Framing the Future of Aviation, 6 March 2015, available at: http://ec.europa.eu/transport/modes/air/news/doc/2015-03-06-drones/ 2015-03-06-riga-declaration-drones.pdf (accessed on 18 April 2017).

166

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

ed that the EU, led by EASA, should immediately develop rules pertaining to the safe operation of UAS.62 Consequently, the European Commission, supported by EASA, is working to amend the scope and competences of the EU through the revision of Regulation 216/2008.63 As it stands at the time of writing this paper, the Draft Proposal Regulation 216/2008 contains three types of risk-based categories for unmanned aircraft, the essences of which can be found in Articles 45–47: 1. ‘open’ is a category of UA operation that, considering the risks involved, does not require a prior authorisation by the competent authority before the operation takes place; 2. ‘specific’ is a category of UA operation that, considering the risks involved, requires an authorisation by the competent authority before the operation takes place and takes into account the mitigation measures identified in an operational risk assessment, except for certain standard scenarios where a declaration by the operator is sufficient; 3. ‘certified’ is a category of UA operation that, considering the risks involved, requires the certification of the UA, a licensed remote pilot and an operator approved by the competent authority, in order to ensure an appropriate level of safety.64

In anticipation of the growing EU competencies, the Agency has proposed an NPA that will contain the specific technical rules on the ‘open’ and ‘specific’ categories. The NPA is a document designed to highlight to stakeholders how the Agency plans to regulate unmanned aircraft. The NPA was open for comments from stakeholders, it will be revised following this input, and then submitted to the European Commission for adoption as a Regulation.65 The proposed rules, however, contain several issues in light of the potential automation of some unmanned aircraft which will be explored below.

62 See ibid. 1. UAS need to be treated as new types of aircraft with proportionate rules based on the risk of each operation; 2. EU rules for the safe provision of UAS services should be developed immediately and the rulemaking process to be led by the Agency; 3. technologies and standards need to be developed for the full integration of UAS in the European airspace; 4. public acceptance is key to the growth of drone services; and 5. the operator of an UAS is ultimately responsible for its use. 63

European Commission (note 35).

64

EASA, Concept of Operations for Drones: A Risk Based Approach to Regulation of Unmanned Aircraft, available at: http://www.easa.europa.eu/system/files/dfu/204696_EASA_concept_drone_ brochure_web.pdf (accessed on 18 April 2017); id., Advance Notice of Proposed Amendment 2015-10 Introduction of a Regulatory Framework for the Operation of Drones, A-NPA 2015-10 (2015). 65 Id., Management Board Decision 18-2015: Replacing Decision 01/2012 concerning the procedure to be applied by the Agency for the issuing of opinions, certification specifications, acceptable means of compliance and guidance material (‘Rulemaking Procedure’), 15 December 2015.

INTERNATIONAL CIVIL AVIATION

167

1. Responsibility One of the core objectives of aviation safety rule-making is to establish a responsible party in the event of an accident or incident. Within this context, responsibility is not always linked to causation, but rather aims to ensure that at least one entity is tasked with safeguarding that the set safety standards are met and, if not, that there is a clearly responsible entity. This approach has also been applied to unmanned aviation, but this has produced a novel issue. Article 3 (1) NPA explicitly states that the operator is responsible for the safe operation, registration, electronic identification, and geo-fencing of the unmanned aircraft: The UAS operator is responsible for the safe operation of the UA. The UAS operator shall comply with the requirements laid down in this Regulation and in other relevant EU and national regulations, in particular those related to security, privacy, data protection, liability, insurance and environmental protection.

This was enshrined in PART-UAS of the NPA which states that the “operator of UA shall have the ultimate responsibility”.66 The responsible party is clearly highlighted, however, it does not define ‘operator’. The definition is provided in Article 3 (h) Regulation 216/2008: “‘operator’ shall mean any legal or natural person, operating or proposing to operate one or more aircraft or one or more aerodromes”. As aviation safety governance exists in a regulatory context, there was the desire to include the unmanned aircraft rules within the existing system for manned aviation. Therefore, the definition of ‘operator’ is compatible with manned aviation as Article 2 (dd) NPA states: “‘unmanned aircraft system (UAS) operator’ means any legal or natural person who operates or intends to operate one or more UAS”. Within the context of manned aviation the circular definition of operator is quite axiomatic as, for example, an aircraft leased to Lufthansa would be deemed to be operated by Lufthansa, since it is the air carrier that conducts the flight. The identification of the operator becomes less clear with unmanned aviation. The above definition leaves it unclear who is the operator in less formal situations than that comparable with commercial manned aviation. Civil unmanned flights in the EU are predomi66 Id., ‘Prototype’ Commission Regulation on Unmanned Aircraft Operations, 22 August 2016, UAS.OPEN.20 and UAS.SPEC.10, available at: https://www.easa.europa.eu/system/files/dfu/UAS% 20Prototype%20Regulation%20final.pdf (accessed on 18 April 2017).

168

GERMAN YEARBOOK OF INTERNATIONAL LAW 60 @ 2017

nantly carried out by private individuals, with no or minimal organisational structure, with small light aircraft. Thus, such operations do not reflect commercial manned aviation operations, such as those performed by Lufthansa, so the operator is unlikely to be an air carrier. Therefore, an operator may be identified as the owner of the unmanned aircraft, the supervisor in case of flights conducted by children, the actual pilot at the time of flight, contractors in case of commercial agreements, or any third party that exercises some control over the operation such as automated technologies. This matter is further complicated as one entity may have several roles, the pilot or supervising entity can change during the course of a flight, or third parties may have varying levels of involvement. It may seem reasonable to conclude that in case of a simple remotely controlled unmanned aircraft owned and flown by one individual this person would be the operator and thus responsible. However, this becomes less clear when a fully automated unmanned aircraft, which just requires mission instructions and to be turned on, then crashes. Here, the owner and user have relatively little control over the aircraft when in flight, as the pilot could then reasonably be seen as the automated system. The automated system may have been developed by a third party, and navigational tools used by the automation technology, such as GPS, are also likely to be those of a third party. This then leaves it unclear who would or should be responsible, a matter which needs also to be balanced based on the levels of automation and control exercised by the individuals involved. This point is then exacerbated with autonomous unmanned aircraft which involve even less direct human intervention.

2. Visual Line of Sight As the setting of fixed limitations was a large point of criticism of Regulation 216/2008 under Annex II which restricted the scope of the Agency’s competencies for unmanned aircraft of 150 kg and above, it was intended to remove such demarcations to allow for a broader and more operation-centric and risk-based approach. However, this proved quite difficult in practice and, instead, EASA maintained and increased the existence of fixed inflexible limitations.

INTERNATIONAL CIVIL AVIATION

169

The NPA, within the context of operational limitations for the ‘open’ category, has prescribed maximum altitude limitations to reduce risks to other air space users and to ensure that the pilot can safely and correctly control the aircraft. These limitations, within each subclass of the ‘open’ category, are: 1. C0: 50 m vertical (Appendix I.1) 2. C1 – C4: