I: Policy, high-level principles, trends and perspectives
Artificial intelligence and machine learning in the financial sector: Legal-methodological challenges of steering towards a regulatory ‘whitebox’
Artificial Intelligence (Al) and Machine Learning (ML) are ubiquitous phenomena. Their use has powerful implications, however Janus-faced. Amongst the interrelated challenges are particularly the degradation of truth and precision surveillance. Technological advances outpace relevant policies as ‘The Global Risks Report 2020’ of the World Economic Forum (WEF Risks Report) states, rightly. In addition, technological rivalries render the future geopolitical scenario uncertain. Data are a key aspect in Al techniques. The data race to foster Al has been unleashed for quite some time with both private and public actors at a global scale. Legal frameworks as to personal and non-personal data10 are crucial, here. The focus has gradually shifted from data governance to Al governance.11 Policy makers at international, European Union (EU) - here particularly the European Commission (Commission) - and national level have acknowledged the need for more general policies in the context of Al.12 International initiatives developing governance standards for Al have emerged.13 Such growing awareness of the urgency of the matter is desirable. But the current disruption of the multinational system fosters fragmentation rather than alignment of the responses to the significant risks.14 The global challenges of Al development, however, require global governance.13 At supranational level such as in the EU, one might observe a first move in this direction: the emergence of Al regulation.16 Legal academia is alert to it. It has started to contribute to the discussion of Al regulation.17
These more general observations in mind, the chapter zooms in on the financial sector. In the financial sector,18 the Financial Stability Board (FSB)19 has dedicated its report ‘Artificial Intelligence and Machine Learning in Financial Services’ (FSB Report Al and ML)20 to the matter quite early. In the FSB’s view, the use of Al and ML is spreading rapidly while assessing it remains constantly provisional due to the paucity of (each) current robust data.21 The pressure for change on actors in the financial sector is huge; their digitalisation strategies might well become a litmus test for whether they serve
- 10 At EU level, the more general starting point is the Charter of Fundamental Rights of the European Union  OJ EU C 326/391 and more specifically inter alia the following regulations: Regulation (EU) 2016/679 of the European Parliament and of the Council of 27 April 2016 on the protection of natural persons with regard to the processing of personal data and on the free movement of such data, and repealing Directive 95/46/EC (General Data Protection Regulation)  OJ EU L 119/1, and Regulation (EU) 2018/1807 of the European Parliament and of the Council of 14 November 2018 on a framework for the free flow of non-personal data in the European Union  OJ EU L 303/59.
- 11 See Ronan Hamon, Henrik Junklewitz and Ignacio Sanchez (n 8) 9.
- 12 See Gudula Deipenbrock, ‘FinTech - Unbearably Lithe or Reasonably Agile? - A Critical Legal Approach from the German Perspective’ (2020) 31 EBLR 3, 30.
- 13 For more information on this, see eg, Peter Cihon, Technical Report- Standards for Al Governance: International Standards to Enable Global Coordination in Al Research & Development (April 2019).
- 14 Compare also eg, World Economic Forum - here Emilio Granados Franco (lead author) et al, WEE Risks Report (n 6) 64, which calls for multilateral cooperation, here.
- 15 For more information on this recommending international standards as tools of Al policy, see eg, Peter Cihon (n 13) 7 ct seq. and 32.
- 16 See also eg, Heinz-Uwe Dettling and Stefan Krüger, ‘Erste Schritte im Recht der Künstlichen Intelligenz - Entwurf der “Ethik-Leitlinien für eine vertrauenswürdige KT” (2019) MMR 211, 217.
- 17 For more information on this, see eg, Thomas Wischmeyer and Timo Rademacher (eds), Regulating Artificial Intelligence (Springer 2020). See also Mario Martini, Blackbox Algorithmus - Grundfragen einer Regulierung Künstlicher Intelligenz (Springer 2019).
- 18 If not expressly indicated otherwise, the term financial sector shall include financial services, financial markets and thereby include both FinTech - see the definitional approach in item 2 - firms and incumbent firms as actors.
- 19 For more information on the FSB as an international body, promoting global financial stability by coordinating the development of regulatory, supervisory and other financial sector policies, and conducting outreach to non-member countries, see FSB
- 20 FSB, Artificial Intelligence and Machine Learning in Financial Services - Market Developments and Financial Stability Implications (1 November 2017).
- 21 FSB, FSB Report Al and ML (n 20) 1 and 3.
Artificial intelligence and machine learning 5 the good of human beings. The use of Al and ML requires close monitoring. Policy and law makers as well as regulators and supervisors worldwide have to be utmost alert to its disruptive potential. The latter includes fundamental technological risks requiring profound Al safety research and work. Value alignment is discussed, here. The same is true for the control problem. A global, concerted, far-sighted, long-term oriented, responsible and sustainable governance of Al and ML is the urgent law of reason. It shall ensure a regulatory ‘whitebox’. The discourse on Al and ML requires rationalisation as observed by experts ranging from moral philosophers to the World Economic Forum in its report ‘The New Physics of Financial Services’ (WEF New Physics Report) with a view to the financial sector. The chapter aims to contribute to such rationalisation. Focus is on the legal-methodological perspective. It is organised as follows. After this introduction, the chapter discusses selected pivotal legal-methodological challenges to be tackled at the outset of any policy, legal, particularly regulatory and supervisory approach to Al and ML in the financial sector. It seeks to address the following questions. How might Al, ML and related phenomena best be defined? How might one best relate Al and ML to law, methodologically? Is it necessary to put (the policy and legal (-political) approach to) the sector-specific use of Al and ML in the financial sector in context with that to the more general use of Al and ML? What challenges are to be tackled to establish the relevant facts of the case? The chapter then concludes. The strict limitations as to the volume of this chapter require a strict selection of issues discussed. Any reference to relevant sources and (legal) literature is also limited to more exemplary sources.
-  All following references to items are references to items of this chapter. All following references to sources use the same writing of titles as to capital letters for reasons of stylistic consistency. 2 See the definitional approaches in item 2.1.1 of this chapter. 3 See the definitional approaches in item 2.1.2. 4 With a view to Al, see eg, De Nederlandsche Bank - here Joost van der Burgt, General Principles for the Use of Artificial Intelligence in the Financial Sector (2019) 33. 5 Eleonore Pauwels, The Neiv Geopolitics of Artificial Intelligence (15 October 2018)
(Call-off date for all hyperlinks, unless stated otherwise: 28 May 2020). 6 See World Economic Forum - here Emilio Granados Franco (lead author) et al - in partnership with Marsh & McLennan and Zurich Insurance Group, The Global Risks Report 2020 (15th edn, no date) 64 et seq. 7 Compare also eg, World Economic Forum - here Emilio Granados Franco (lead author) et al, WEF Risks Report (n 6) 14 et seq. 8 See eg, Ronan Hamon, Henrik Junklewitz and Ignacio Sanchez, JRC Technical Report - Robustness and Explainability of Artificial Intelligence (EUR 30040 EN, Publications Office of the European Union 2020) 7. 9 See already Gudula Deipenbrock, ‘Is the Law Ready to Face the Progressing Digital Revolution? -General Policy Issues and Selected Aspects in the Realm of Financial Markets from the International, European Union and German Perspective’ (2019) 118 ZVglRWiss 285, 293.
-  For more information on the discussion in the realm of ethics, see eg, Julian Nida-Rümelin and Nathalie Weidenfeld, Digitaler Humanismus - Eine Ethik für das Zeitalter der Künstlichen Intelligenz (Piper 2018) 77 et seq. 2 See from the financial stability perspective eg, FSB, FSB Report AI and ML (n 20) 1. 3 See in this context also Max Tegmark, 'Let’s Aspire to More than Making Ourselves Obsolete’ in John Brockman (ed), Possible Minds - Twenty-Five Ways of Looking at Al (Penguin Press 2019) 76,81 et seq., who argues that the controversial position on Al safety research is no longer to advocate for it but to dismiss it. 4 See eg, Tom Griffiths, ‘The Artificial Use of Human Beings’ in John Brockman (ed), Possible Minds -Twenty-Five Ways of Looking at Al (Penguin Press 2019) 125, 128 etseq. 5 See eg, Stuart Russell, Human Compatible: Artificial Intelligence and the Problem of Control (Penguin Random House 2019). 6 See eg, Julian Nida-Rümelin, ‘Preface* in Julian Nida-Rümelin and Nathalie Weidenfeld, Digitaler Humanismus - Eine Ethik für das Zeitalter der Künstlichen Intelligenz (Piper 2018) 9, 11, who argues that there is beyond apocalyptic disaster scenarios and ‘technicism-based’ hopes of salvation a middle course of maintaining and improving the human living conditions through culturally, socially and politically controlled use of technological means. 7 See World Economic Forum - here R. Jesse McWaters (lead author) et al - prepared in collaboration with Deloitte, The New Physics of Financial Services - Understanding how Artificial Intelligence is Transforming the Financial Ecosystem (August 2018) 9, stating that the public discourse on Al in financial services is highly sensationalised, creating an excess of both exuberance and fear. Please see the disclaimer (World Economic Forum - here R. Jesse McWaters (lead author) et al, WEE New Physics Report, 1).