As regulators and banks move from broad AI principles to more practical controls for generative and mediated AI, Singapore’s central bank has released an industry-developed toolkit to help financial institutions manage risks from AI.
Monetary Authority of Singapore (MAS) said it has completed the second phase of the project MindForge Project with the release of an AI Risk Management Toolkit for the financial services industry.
The toolkit was developed by a consortium of 24 banks, insurers, capital markets companies and other industry players and aims to support risk management across traditional AI, generative AI and emerging agency AI systems.
The core of the package is an “AI Risk Management Operationalization Handbook” that provides financial institutions with practical guidance on how to implement AI risk controls.
MAS said the handbook is accompanied by an appendix that compiles case studies from financial institutions, outlining lessons learned, implementation challenges and approaches to managing AI in different organizational settings.
The handbook is structured around four areas that reflect MAS’s proposed AI Risk Management Guidelines: scope and oversight, AI risk management, AI lifecycle management, and organizational enablers such as capabilities, infrastructure, and resources.
MAS will launch a public consultation on the proposed guidelines in November 2025, saying they will apply across the financial sector and set out supervisory expectations for oversight, governance and lifecycle controls for the use of AI.
MAS said it was still reviewing feedback from that consultation. The central bank added that the handbook will be updated periodically as the use of artificial intelligence in industry matures and supervisory expectations evolve.
To support broader adoption, MAS plans to establish an AI risk management working group under the BuildFin.ai initiative and plans to bring together MindForge consortium members and other practitioners to develop implementation resources and frameworks for newer technologies such as agency AI.
The MindForge Project was launched in mid-2023 as part of MAS’ broader effort to promote the responsible adoption of AI in finance.
MAS is also building an AI agenda across wider finance, including a 2024 factsheet on good practice in AI and generative AI model risk management, and a 2025 partnership with the UK Financial Conduct Authority on AI in financial services.
MAS chief fintech officer Kenneth Gay said the toolset represents a “major step forward” in promoting responsible use of AI in finance and will help strengthen governance and risk management practices across the sector.
MAS is trying to move the market from the principle setting stage to implementation. This is important because many banks already have AI policies on paper, but generative and mediated AI creates newer operational risks around oversight, accountability, model behavior, and lifecycle controls.
MAS appears to be signaling a more collaborative supervisory approach by publishing a handbook created with industry participants rather than issuing rules alone.
