Draft Guideline on the Use of Artificial Intelligence in the Financial Sector
On July 3, 2025, the Autorité des marchés financiers (AMF) released a French-only draft guideline focusing on the use of artificial intelligence (AI) in the financial sector. This guideline applies to authorized insurers, financial services cooperatives, authorized trust companies, and other authorized deposit-taking institutions. It outlines the AMF’s expectations for managing risks related to AI systems and ensuring fair treatment of clients. Stakeholders are invited to submit comments by November 7, 2025.
Definition of AI System
The guideline defines an AI system as a machine-based system that infers from inputs to generate outputs such as predictions, content, recommendations, or decisions. These outputs can influence physical or virtual environments. The guideline also notes that AI systems differ in autonomy and adaptiveness after deployment.
Key Expectations
Risk Rating
Financial institutions must assign a risk rating to each AI system and review these ratings and their factors periodically, at least once a year. The risk rating guides the scope of policies, processes, and procedures for managing the AI system’s lifecycle, governance, risks, and consumer treatment. Factors to consider include data characteristics, controls, and overall exposure risk. A provisional rating should be assigned initially and refined as more data becomes available.
Lifecycle of AI Systems
Institutions are expected to develop and document processes for each phase of an AI system’s lifecycle based on its risk rating. Higher-risk AI systems require more frequent monitoring and corrective actions. The lifecycle is divided into three phases:
- Design or Acquisition Phase: Evaluate training data quality and govern the design or acquisition process.
- Use and Monitoring Phase: Conduct validations, internal audits, and continuous supervision of performance and use.
- Modification or Decommissioning Phase: Manage updates or retirement of the AI system effectively.
Decisions to implement AI over alternative solutions must be documented, demonstrating the chosen system’s appropriateness relative to its risk.
Governance
Clear policies and procedures must define roles and responsibilities for all stakeholders across the AI system lifecycle. Each AI system requires a designated AI manager accountable throughout its lifecycle, reporting to senior management. Those setting policies should have strong knowledge of AI risks, the institution’s risk appetite, and ethical standards. Operators should understand AI and its risks practically.
The guideline highlights four key stakeholder groups:
- Board of Directors
- Senior Management
- Risk Management
- Internal Auditors
The board must collectively understand AI sufficiently to make informed decisions about evaluation, deployment, and risk management.
Risk Management
Institutions should adopt policies and procedures proportional to their size, operations, and risk profile. They should identify major AI-related risks, maintain an updated risk taxonomy, and regularly assess risk appetite and tolerance. A centralized directory of all AI systems must be maintained to support transparent supervision and risk control across the institution.
Fair Treatment of Customers
The guideline adds expectations on fair customer treatment in AI use. Financial institutions must ensure their ethics codes address AI-specific challenges. They need to identify, document, and report any discriminatory or biased outcomes from AI decisions, escalating issues to senior management.
Regarding customer consent and communication, institutions must clearly explain how personal data will be used, including linking with secondary data that may affect data quality. Customers should receive simple, clear explanations when decisions are made or assisted by AI systems, whether autonomously or with human oversight.
Significance of the Guideline
This draft guideline represents a major step in regulating AI within Québec’s financial sector. It emphasizes the need for institutions to align with evolving standards to reduce operational, reputational, and legal risks.
For financial professionals looking to enhance their understanding of AI implementation and governance, exploring targeted training can be beneficial. Relevant courses and certifications are available at Complete AI Training.
Your membership also unlocks:
