Monetary Authority of Singapore releases AI risk rules

According to the paper published on November 13, all regulated financial institutions, including banks, insurers, and market intermediaries, should properly govern and control the use of artificial intelligence.

The guidelines cover the full range of AI applications in finance, from customer-facing tools and internal analytics to autonomous systems handling complex processes. They apply across the sector to banks, insurers, capital markets participants, and other license holders supervised by MAS.

MAS set supervisory expectations for board oversight, risk frameworks, and lifecycle controls in managing AI systems, including generative models and AI agents. The regulator said the standards are principles-based and proportionate, allowing requirements to scale based on an institution’s size and the risk level of each AI use case.

Among current risks, MAS warned that inaccurate AI risk assessments could lead to significant financial losses, while unexpected system behavior might disrupt key operations or harm customers. It also highlighted concerns around generative AI, such as hallucinations and unpredictable behavior.

Other issues flagged include security threats like prompt injection attacks, data privacy risks from using third-party services, and possible copyright violations when AI models are trained on protected material. MAS also cautioned against heavy dependence on a few dominant AI providers and the growing tendency for workers to rely too much on AI tools in their daily roles.

To address workforce challenges, Singapore’s three major banks are retraining 35,000 local employees over the next one to two years to adapt to AI-related changes in their roles and processes. MAS deputy managing director Ho Hern Shin said the proposed guidelines provide “clear supervisory expectations to support financial institutions in leveraging AI,” while maintaining safeguards against emerging risks. Stakeholders can submit feedback on the proposals until January 31, 2026.

In October, the Financial Stability Board (FSB) also warned that wider AI use in finance increases risks – particularly concentration among a few model and infrastructure providers – and urged regulators to strengthen oversight of outsourced tools and share incident data.