MAS consults on guidelines for AI risk management
Singapore's central bank has released a consultation paper with proposed guidelines for artificial intelligence risk management across the financial sector. The rules would apply to all financial institutions and cover a wide range of AI uses, including generative AI and newer setups like AI agents.
"MAS recognises that AI can be applied to a wide range of use cases, and that the risks associated with different usage of AI may vary based on the scale, scope and business models of FIs," says the regulator. The guidelines set expectations that are proportionate to the size of the firm, how AI is used, and the risk profile.
Who this affects
All financial institutions operating in Singapore. If you're using AI in any part of the business-from underwriting to chatbots to agentic automation-these expectations will matter.
Governance: leadership is on the hook
Board and senior management are expected to own AI risk. That includes establishing frameworks, structures, policies, and processes, and setting the right risk culture for AI adoption.
Translation: assign clear accountability, define decision rights, and make sure AI risk is discussed at the top table, not buried in an innovation team.
Know your AI footprint
FIs should implement firm-wide identification processes for AI use and maintain an accurate, up-to-date AI inventory. No inventory, no control.
Risk materiality assessments should consider three lenses: impact, complexity, and reliance. This creates a simple triage: what needs rigorous controls now, what needs basic safeguards, and what can wait.
Controls MAS expects (apply based on materiality)
- Data management and quality
- Fairness and bias mitigation
- Transparency and explainability
- Human oversight and accountability
- Third-party and vendor risks
- Evaluation and testing (pre-deployment and ongoing)
- Monitoring and change management
"These proportionate, risk-based guidelines enable responsible innovation by financial institutions that implement the relevant safeguards to address key AI-related risks," says MAS's deputy managing director, Ho Hern Shin.
Complementary industry handbook
To support implementation, the Project MindForge consortium will publish an "AI Risk Management Executive Handbook." It outlines key components of strong AI risk management, with a more detailed, practical document to follow next year.
"The guidelines and the handbook will work together. For example, the Guidelines will require risk materiality assessments, while the Handbook will provide examples of how such assessments are done," says MAS managing director Chia Der Jiun at the Singapore FinTech Festival 2025.
What leaders should do now
- Appoint a senior accountable executive for AI risk and align board oversight.
- Build an enterprise AI inventory within 30-60 days; define what counts as "AI."
- Stand up a simple materiality scoring model (impact, complexity, reliance) and triage all use cases.
- Codify controls: data lineage and quality checks, bias testing, explainability standards, human-in-the-loop thresholds.
- Map third-party AI dependencies; update contracts for model changes, monitoring, and audit rights.
- Require pre-deployment testing, ongoing monitoring, and issue escalation playbooks.
- Run limited-scope pilots for higher-risk AI while controls mature.
- Upskill your teams on AI governance and practical usage so policies stick in day-to-day work.
If you need a curated place to start training plans for finance teams, see our overview of AI tools for finance.
Timeline and how to respond
The consultation paper is available on the MAS website. Interested parties should submit feedback by Jan 31, 2026.
Access MAS consultations here: MAS Consultations. For event context, see the Singapore FinTech Festival.
Bottom line: treat this as your blueprint. Get your governance in place, quantify materiality, and install practical controls. That's how you scale AI use without adding surprise risk.
Your membership also unlocks:
