OSFI has finalised an enterprise-wide, risk-based model-risk guideline that expressly includes AI and machine-learning models across federally regulated financial institutions.
Guideline E-23 broadens model inventories and lifecycle governance, including expectations for data, explainability, independent review, deployment and third-party models.
Federally regulated banks, foreign bank branches, insurers, trust and loan companies; proportionate to risk and institutional complexity.
Summarised from the primary instruments below. Open the source before relying on an obligation for legal or compliance decisions.
Maintain an enterprise model-risk framework, risk ratings, inventory and clear lifecycle responsibilities.
Independently review conceptual soundness, performance, explainability and limitations proportionate to risk rating.
Use accurate, representative, traceable and compliant data with documented provenance and controls for synthetic or proxy data.
The original authority remains the source of truth. Dates below are publication and effective dates, not OneBench ingest dates.
Office of the Superintendent of Financial Institutions
Final enterprise-wide model-risk guideline expressly covering AI and ML models across regulated banks and insurers.