MAS Signals Third-Party AI Failures Won't Shield Singaporean Fintechs

The Monetary Authority of Singapore (MAS) is drawing a hard line on AI liability, making it clear that relying on third-party vendors will not exempt financial institutions (FIs) from regulatory accountability. According to reporting from The Register, the MAS published its first Guidelines on Artificial Intelligence Risk Management for Financial Institutions on Wednesday, which explicitly state that FIs remain accountable for the AI used in their services, even when that technology is developed or operated by third parties.

MAS warns that the probabilistic nature of AI—and the even greater complexity of generative AI—increases the risk of biased behavior and unexpected outcomes. To mitigate this, the regulator is requiring all AI use cases to undergo independent reviews by parties not involved in the development process before they are deployed into production. Additionally, FIs must conduct technology and cybersecurity reviews to ensure secure deployment.

Because AI can suffer from performance degradation, model staleness, and data drift, MAS is mandating ongoing monitoring of both internal and third-party AI systems. The regulator expects FIs to obtain sufficient assurance from vendors and implement compensating controls to bridge any assurance gaps. If risks exceed an institution's risk appetite, MAS suggests they limit, suspend, or replace the third-party service.

Further requirements include maintaining up-to-date inventories of all AI used in business operations and developing contingency plans—such as manual processes or alternative systems—for high-risk applications to ensure business continuity during AI failures. These guidelines are set to come into force on October 7, 2027.