Singapore MAS Guidelines on Artificial Intelligence Risk Management (2026)

MAS's supervisory expectations for AI risk management in Singapore financial institutions, effective from 7 October 2027: proportionate application with a basic governance route for low-impact use, board and senior management oversight with AI in the risk appetite, a designated control function for AI identification, an AI inventory and a risk materiality assessment on impact, complexity and reliance, life cycle controls covering data management, transparency, fairness, human oversight, third-party AI, selection, evaluation and testing, technology and cyber security, reproducibility, independent pre-deployment validation, monitoring with drift checks and kill switches, change management and decommissioning, and the people and infrastructure capability to support it. It comprises 30 controls organised across 6 domains, and applies in Singapore.