AI Risk Management in Singapore
AI risk management in Singapore
How Singapore's approach evolved, from principles to guidelines to a working toolkit. Each item links to the source, with a short note on what it is. I wrote the AIRG and led the review behind the 2024 paper.
-
2018 MASFEAT Principles
MAS's founding principles for responsible AI and data analytics in finance: fairness, ethics, accountability and transparency. Among the first such guidance from a financial regulator, and the origin of the 2025 guidelines.
-
2019 IMDAModel AI Governance Framework
Singapore's cross-sector, voluntary framework for deploying AI responsibly, from IMDA and the PDPC. A second edition followed in 2020.
-
2019 to 2023 MASVeritas Initiative
A MAS-led consortium that turned the FEAT principles into measurable assessment methodologies and open-source tools.
-
2022 IMDAAI Verify (testing framework and toolkit)
IMDA's AI governance testing framework and software toolkit: technical tests for fairness, explainability and robustness, plus process checks against the governance principles. Launched as the world's first such toolkit.
-
2023 IMDAAI Verify Foundation
A non-profit set up by IMDA with global technology partners to steward AI Verify as open source and help shape international AI testing standards.
-
2024 IMDAModel AI Governance Framework for Generative AI
IMDA and the AI Verify Foundation's extension of the framework to generative AI, adding dimensions such as content provenance, security and testing.
-
2024 MASAI Model Risk Management (Information Paper)
MAS's observations from a thematic review of banks' AI model risk practices, and the operational groundwork for the AIRG. I led the thematic review behind it.
-
2025 MASGuidelines on AI Risk Management (AIRG)
MAS's proposed supervisory expectations for AI risk management across all financial institutions: oversight, inventory, materiality, lifecycle controls, and capability. Issued for consultation. I developed it. See the AIRG explained and the AIRG in practice.
-
2026 IMDAModel AI Governance Framework for Agentic AI
IMDA and the AI Verify Foundation's framework for agentic AI, structured around bounding the risks, meaningful human accountability, technical controls, and end-user responsibility.
-
2026 MASAI Risk Management Toolkit (Project MindForge)
A practical toolkit from MAS and industry partners, with an operationalisation handbook and case studies to help firms put AI risk management into practice.
-
2026 MASSafeguards for Agentic Finance at Runtime (SAFR)
MAS's work on runtime safeguards for agentic AI in finance. I contributed to it. See governing agentic AI.
Two tracks run in parallel. The financial-sector line from MAS: FEAT, Veritas, the 2024 Information Paper, the AIRG, the MindForge Toolkit and SAFR. And the national and cross-sector line, led by MDDI and IMDA with the AI Verify Foundation: the Model AI Governance Frameworks, and AI Verify. If you are a financial institution, the MAS line is the one you are supervised against.
For what all of this asks of you in practice, see AI risk management and the AIRG in practice. I train and advise Singapore's financial institutions on all of it; see the courses and workshops, or get in touch.