As artificial intelligence (AI) increasingly becomes a strategic driver for Canadian financial institutions, the regulatory framework governing AI use was also reinforced in 2025. One important addition is the Autorité des marchés financiers (AMF) publishing, in July 2025, a draft Guideline for the Use of Artificial Intelligence, which outlines the expectations applicable to financial institutions regarding the responsible use of artificial intelligence systems (AIS).
This guideline provides financial institutions operating in Québec with a clearer idea of the AMF’s expectations regarding AIS use. A consultation and comments period was open from publication until November 2025 and the guideline will, once finalized, complement the Office of the Superintendent of Financial Institutions (OSFI)’s Guideline B-13, in effect since January 2024, which governs technology and cyber risk management for federally regulated financial institutions.
Although distinct in their approaches, the AMF draft guideline and OSFI’s B-13 Guideline pursue converging objectives and will need to be applied in parallel by financial sector institutions operating both in Québec and under federal regulation. This coexistence raises a practical challenge: how can these requirements be harmonized to avoid duplicating efforts and ensure consistent implementation?
This article aims at addressing this challenge by comparing the two regulatory frameworks, highlighting areas of convergence and key integration levers to help institutions establish unified and effective technology governance.
Strategic alignment
While the AMF guideline specifically targets AIS, it builds on the broader technology risk management principles set out by OSFI. Both guidelines are grounded in a risk-based approach, establish clear governance expectations and place a strong emphasis on operational resilience.
Financial institutions can leverage the governance structures, technology validation processes and ongoing monitoring mechanisms already implemented under Guideline B-13 to address many of the AMF’s expectations.
Leveraging points of convergence
To avoid duplicating efforts, financial institutions can capitalize on the many overlaps between the two guidelines. The roles and responsibilities outlined in OSFI’s Guideline B-13 for technology governance can be extended to AIS, notably by assigning an individual to be responsible for each AIS and ensuring clear lines of accountability up to senior management.
Similarly, the technology lifecycle management processes set out in Guideline B-13 (including system development lifecycle practices) can be adapted to incorporate AIS-specific lifecycle stages such as design, validation and ongoing monitoring, including by integrating performance metrics and bias detection mechanisms.
The technology asset inventory required under Guideline B-13 can also be expanded to include the AIS-specific data the AMF requires institutions to track, such as risk ratings, training data and explainability features. Technology incident management protocols can also be expanded to cover risks specific to AI, such as model drift, discrimination and bias, hallucinations, or intellectual property infringements, in addition to risks already being tracked in accordance with Guideline B-13 requirements.
Data protection is also an obvious point of convergence. OSFI addresses this issue from the perspective of information security, confidentiality and availability, while the AMF sets out additional requirements, particularly regarding data quality and limitations on data use. Internal data protection policies can therefore be harmonized into a single coherent set to cover the specific risks associated with AIS, while building on the technical foundations already in place in accordance with OSFI requirements.
The AMF’s distinctive contributions
The AMF’s guideline imposes a set of new requirements specifically tailored to the challenges associated with using AI in financial services.
The guideline provides for a modulation of obligations according to the risk rating assigned to each AIS, which financial institutions will be required to assign to their AIS. This rating will serve as the foundation for adopting internal policies, procedures and controls over such AIS, in addition to influencing validation, documentation and supervision requirements imposed on the financial institution.
It also emphasizes fair treatment of clients and transparency toward them, in particular by detecting biases and explaining automated decisions. Finally, it incorporates an environmental dimension by requiring greenhouse gas emissions associated with AIS be factored into the risk analysis.
Although these requirements are new, in many cases they can be integrated into the governance, risk management and technology oversight mechanisms already in place under Guideline B 13, provided their scope is broadened to address AI-specific considerations.
Meeting both AMF and OSFI requirements
Rather than multiplying processes, financial institutions may view this as a strategic opportunity to design an internal compliance framework that meets both the requirements of OSFI’s Guideline B-13 and the AMF’s AI expectations. This possibility is based on similarities observed in the regulatory approaches of the two regulatory bodies, particularly for governance, risk management and technology oversight. Such an approach will promote more efficient implementation of this framework for financial institutions subject to the requirements of both bodies, while strengthening the consistency and robustness of technology governance mechanisms. With the AMF’s consultation period regarding the draft guideline on AIS now closed, we can expect a final version to soon be available. In 2026, financial institutions subject to AMF requirements need to ensure their compliance with AMF guidelines and should further consider how to integrate these new requirements in their existing controls to streamline operations and effectiveness. It will also be interesting to see whether any other provincial regulators issue similar guidelines, which may further contribute to the challenge of multijurisdictional compliance.
The authors would like to thank Charles-Antoine Bordeleau, articling student, for his contribution to preparing this legal update.
Volume 1: Looking back on key insights from 2025