The AI regulatory framework for French financial services
French financial institutions operate in a particularly dense AI regulatory environment. The ACPR and the AMF apply their own sectoral frameworks, which sit on top of the obligations of the EU AI Act and the GDPR. This multiplicity of reference frameworks calls for an integrated approach to AI governance rather than siloed regulatory responses.
The ACPR's expectations on AI models
The ACPR has published several guidance documents on the use of AI in the banking and insurance sector. That guidance draws on the work of the FSB (Financial Stability Board) and the EBA (European Banking Authority) on model risk management. Institutions are expected to have: an inventory of the AI models used in critical decision-making processes, documentation of the performance and the perimeter of validity of each model, independent validation procedures, and continuous monitoring of performance drift.
For insurers, the ACPR pays particular attention to AI pricing models, notably the risk of indirect discrimination through proxy variables, and to transparency towards policyholders about the factors influencing their premium.
Credit scoring and cumulative obligations
Credit scoring models using machine learning perfectly illustrate the complexity of the regulatory landscape. They are potentially subject simultaneously to: the "high-risk" classification of the EU AI Act (Annex III, point 5), Article 22 of the GDPR and the right to explanation, the ACPR's guidance on model risk management, and the non-discrimination rules of consumer law. Effective governance must respond to these four reference frameworks in a coherent way.
