On Thursday 11 June 2020, the French Prudential Supervision and Resolution Authority (Autorité de contrôle prudentiel et de résolution, or ACPR) published a discussion document on the governance of artificial intelligence (AI) algorithms in Finance. In such publication, the ACPR sets forth a framework for evaluating AI algorithms and the system that should accompany their use and their integration by financial sector players. This document is subject to public consultation until 4 September 2020, and is a unique opportunity for players to actively contribute to the definition of a relevant framework in France for AI in Finance.

Artificial intelligence (AI) is the subject of a number of regulatory discussions, both in Europe and in France. In February 2020, the European Commission published a white paper on AI, subject to public consultation until May 2020^[1]. Its purpose was, in particular, to define a European framework enabling broader use of AI, taking into account related risks.

It is now the turn of the ACPR to solicit stakeholders on the governance of AI algorithms in the financial sector. To this end, it published a discussion paper^[2] on 11 June 2020, which is open for public consultation until 4 September 2020.

This document is part of the work on AI initiated by the ACPR since 2018^[3], notably through its Fintech and Innovation Hub. ACPR’s work to date has enabled it to identify two key issues, the explicability and governance of AI, in three separate areas: combating money laundering and terrorist financing (AML-CFT), internal models, and customer protection.

Within this framework, the ACPR intends to respond to requests for regulatory clarification from financial players regarding these new technologies. It thus sets forth a framework that looks to strike a balance between freedom of innovation, regulatory compliance and reasoned risk management^[4]. In defining this balance, the ACPR identifies two main points: evaluating AI algorithms, and their governance;

With respect to their evaluation, the ACPR identifies four objectives that it describes as “interdependent”^[5], and that it believes should be taken into account in the design and development of AI algorithms:

– the appropriate management of the data collected;

– the algorithm performance;

– their stability (i.e. the robustness and resilience of their behaviour throughout their life cycle); and

– their explainability (i.e. their transparency and auditability).

With respect to their governance, the ACPR stresses the important paradigm shift brought about by the use of algorithms in Finance and the need to re-examine the validation of decision-making processes. It therefore recommends that stakeholders take particular attention in determining the governance rules applicable to algorithms, taking into account the proper integration of algorithms into business processes, their security and their validation process. According to the ACPR, governance issues should be taken into account “as early as the algorithm design phase“^[6], particularly with regard to the integration of AI into traditional processes, the impact of this integration on internal control, the relevance of outsourcing part of the design or operation, and lastly, the internal or external audit functions.

With this new consultation, the ACPR submits to stakeholders its vision as regards the requirements of AI tools that are likely to develop in the financial field. This consultation, open until 4 September 2020, is an opportunity for stakeholders to contribute to the discussions to support, refine or complete the avenues proposed. The definition of a regulatory framework suited to AI uses will enable France to establish itself as a leading nation in this innovative field.
^[1] European Commission White paper on Artificial Intelligence – A European approach to excellence and trust, 19 February 2020.

^[2] ACPR, Document de réflexion – Gouvernance des algorithmes d’intelligence artificielle dans le secteur financier, 11 June 2020.

^[3] ACPR, Intelligence artificielle : enjeux pour le secteur financier, December 2018 ; ACPR, Intelligence artificielle : enjeux pour le secteur financier, Réponses à la consultation publique sur le document de réflexion de l’ACPR de décembre 2018, Analysis of responses, June 2019.

^[4] ACPR, Document de réflexion – Gouvernance des algorithmes d’intelligence artificielle dans le secteur financier, 11 June 2020, p. 5.

^[5] ACPR, Document de réflexion – Gouvernance des algorithmes d’intelligence artificielle dans le secteur financier, 11 June 2020, p. 3.

^[6] ACPR, Document de réflexion – Gouvernance des algorithmes d’intelligence artificielle dans le secteur financier, 11 June 2020, p. 20.