MAS finalises AI risk management guidelines; first expectations apply from October 2027

Implications

Facts

The Monetary Authority of Singapore (MAS) issued its Guidelines on Artificial Intelligence (AI) Risk Management on 7 October 2026, after a public consultation launched in November 2025. The guidelines apply to all financial institutions and to all AI technologies, and each institution decides how to meet them according to the nature and scale of its AI use and the materiality of the risks.

The guidelines formally take effect on 7 October 2027, when the expectations in Sections 3 and 4 (oversight and risk management) apply; the expectations in Sections 5 and 6 (life-cycle controls and capabilities) apply from 7 October 2028. Institutions are expected to assign clear board and senior-management accountability for AI risk, keep an inventory of AI systems, assess risk materiality at both enterprise and use-case level, and apply controls across the AI life cycle, including data governance, testing, human oversight, monitoring and change management. Firms remain accountable for AI supplied by third parties and should consider limiting, suspending or replacing a service whose risks exceed their risk appetite. Lower-risk applications may be governed by simpler policies, and dedicated AI committees are not required.

According to OpenGov Asia, MAS plans to consult the industry in 2027 on whether agentic AI needs further guidance. MAS Deputy Managing Director Ho Hern Shin said greater regulatory clarity lets institutions "innovate with confidence".

Analysis

Hong Kong took a narrower, conduct-focused route earlier. The HKMA's circular of 19 August 2024 on consumer protection in the use of generative AI applies to customer-facing uses by banks and stresses board accountability, a human in the loop, the option to opt out of AI-driven processes or request human review, and disclosure to customers. Singapore's final guidelines are broader in scope, covering every AI technology and every type of financial institution, and focus on firm-wide risk management rather than consumer-facing rules alone.

The timing matters for lenders because adoption is moving ahead of the rules. In a vendor survey published in September, 83% of Singapore respondents said they trust AI to compare loans and 49% would let an AI agent apply for credit on their behalf. ¹ Credit scoring, fraud screening and collections prioritisation are likely to rank as material use cases under a materiality assessment, because they directly affect customer outcomes. The guidelines leave the materiality judgement to each institution rather than listing high-risk uses.

Counterpoints and uncertainties

The guidelines are principles-based and proportionate, so the practical burden will depend on how MAS supervisors interpret materiality in inspections, which the release does not specify. Guidance on agentic AI is still to come in 2027 and could add requirements for AI agents that act for customers. The consumer survey cited is a vendor study and does not measure actual AI use in lending decisions.