The ACPR sets its AI position through discussion documents rather than binding rules. Its June 2020 discussion document 'Governance of Artificial Intelligence in Finance' proposed four evaluation criteria (data management, performance, stability, explainability) and governance principles from design to audit, and on 1 July 2026 it launched a public consultation on algorithmic fairness open until 30 September 2026. In a 1 July 2026 market meeting presentation, the ACPR said that its designation as the AI Act market surveillance authority for finance-specific use cases was proposed in the DDADUE bill and still had to be validated by Parliament.
| Full name | Autorité de contrôle prudentiel et de résolution (ACPR), France's prudential supervision and resolution authority, attached to the Banque de France |
| Role | French prudential supervisor and AML authority with an innovation hub that publishes AI governance and fairness papers and is preparing for AI Act market surveillance in finance |
| Force on banks | Supervisory guidance |
| Applies to | French banks, insurers and other entities under ACPR supervision; the 2020 discussion document covers AI (mainly machine learning) in AML-CFT, internal models such as credit scoring, and customer protection |
| Key document | ACPR discussion document 'Governance of Artificial Intelligence in Finance' (June 2020) |
| Latest move | 1 July 2026: the ACPR launched a public consultation on algorithmic fairness in the financial sector (open until 30 September 2026) and held a market meeting on AI regulation and supervision |
| Documents tracked | 1 · all documents → |
The ACPR's Fintech-Innovation Hub has worked on AI since 2018, running a first public consultation in December 2018 and exploratory workshops from March 2019 on AML-CFT, credit scoring and customer protection, which produced the 2020 discussion document. Its four-level explanation scale (observation, justification, approximation, replication) and its audit approach combining analysis of code and data with empirical testing through challenger models and benchmarking datasets are the most cited parts of that document.
For the EU AI Act, the ACPR says it expects to be the market surveillance authority for financial services and has set up a cross-functional task force on legal issues, AI auditing, European and national coordination, internal organisation and sector communication; an April 2025 working meeting brought together about 50 AI specialists from around 20 institutions. Its July 2026 consultation on algorithmic fairness, covering the legal framework, fairness metrics and bias identification, is part of building an evaluation methodology for AI systems in finance.
What has the ACPR actually published on AI?
| Date | Document | Status |
|---|---|---|
| Jun 11, 2020 | ACPR discussion document on AI governance — Governance of Artificial Intelligence in Finance | Final |
| Date | Type | Document / event |
|---|---|---|
| Sep 30, 2026 | Milestone | ACPR algorithmic fairness consultation closes. Deadline for responses to the ACPR's public consultation on algorithmic fairness in the financial sector. |
| Jul 1, 2026 | Milestone | ACPR launches public consultation on algorithmic fairness. The ACPR published a reflection document on algorithmic fairness in the financial sector and a consultation questionnaire, covering the legal framework, fairness metrics, bias identification methods and practical implementation guidance, with responses invited until 30 September 2026. |
| Apr 2, 2025 | Milestone | ACPR working meeting with AI specialists from about 20 financial institutions. The ACPR held a working meeting with about 50 AI specialists from about 20 institutions to gather feedback on AI Act readiness, industry questions and difficulties, and to identify audit methodology partners. |
| Jun 11, 2020 | Consultation | ACPR discussion document on AI governance — Governance of Artificial Intelligence in Finance. The ACPR's discussion document 'Governance of Artificial Intelligence in Finance', published in June 2020 (dated 11 June 2020 on the ACPR site) by its Fintech-Innovation Hub, proposes how to evaluate and govern AI algorithms in finance. source ↗ |
- Parliamentary validation of the ACPR's designation as AI Act market surveillance authority for finance-specific use cases, proposed in the DDADUE bill
- Publication of consultation responses and any follow-up guidance or evaluation methodology on algorithmic fairness after the 30 September 2026 deadline
- ACPR's AI evaluation methodology intended to become a reference framework for institutions and the supervisor
- Alignment with the AI Act timetable for Annex III high-risk systems, which the ACPR's July 2026 presentation describes as applying from 2 December 2027 under the final omnibus compromise
Does the ACPR have binding rules on AI?
No. Its AI publications are discussion documents and consultations. The June 2020 'Governance of Artificial Intelligence in Finance' paper invited feedback on its proposals, and the 2026 algorithmic fairness paper is likewise a reflection document under consultation.
What are the ACPR's four criteria for evaluating AI algorithms?
Appropriate data management (including fairness and absence of discriminatory bias), performance, stability (robustness over the lifecycle, including model drift) and explainability, for which the ACPR defines four levels of explanation: observation, justification, approximation and replication.
Is the ACPR the AI Act supervisor for French banks?
It expects to be. The ACPR's own 1 July 2026 presentation says its designation as market surveillance authority for finance-specific use cases was proposed in the DDADUE bill but still had to be validated by Parliament; check the current status before relying on it.
Follow every move these regulators make
when one of these regulators moves, the next morning's brief says so · six sourced stories · 7 am ET · free
plus every tracker, bank and agent page update, the morning after · leave any morning