AI Risk Management Guidelines for Banks

Supervisory expectations document for AI model risk management in the financial sector, formally constraining how banks use AI.

MAS released the AI Risk Management Guidelines in December 2024, codifying years of practical experience from FEAT, Veritas, and MindForge into formal supervisory expectations. Coverage includes: model governance (data, training, validation, deployment), third-party AI risk (cloud providers, model vendors, APIs), model monitoring (drift, bias, performance), human-in-the-loop, and incident response and accountability. The accompanying BuildFin.ai platform enables regulated institutions to test and report on a continuous basis. These are among the world's first dedicated supervisory documents for AI in banking, landing ahead of the EU AI Act's financial provisions.