Project MindForge by MAS sets AI risk governance benchmark for financial services
Covers governance, risk identification, and lifecycle controls for AI in financial services.
The Monetary Authority of Singapore (MAS) has completed Phase 2 of Project MindForge, resulting in the release of an AI Risk Management Toolkit developed with 24 financial institutions. The toolkit provides structured guidance for managing risks across traditional AI, generative AI, and agentic AI use cases within the financial sector.
Key Highlights
Industry-wide collaboration across financial institutions: A consortium of 24 banks, insurers, and capital market participants contributed to the toolkit, reflecting practical implementation experiences across multiple AI deployment environments.
Operationalisation Handbook aligned with regulatory direction: The toolkit includes a detailed handbook structured around governance, risk identification, lifecycle controls, and organisational enablers, aligned with MAS’ proposed AI risk management guidelines.
Inclusion of real-world case studies and implementation learnings: A supplementary section compiles institutional use cases, outlining challenges, applied approaches, and risk mitigation practices across different AI adoption contexts.
Coverage across emerging AI technologies including agentic AI: The framework extends beyond traditional models to include generative and agentic AI, addressing newer risk categories linked to autonomy, decision-making, and system interactions.
Ongoing updates and ecosystem capability-building initiatives: The toolkit will be periodically updated, supported by a dedicated industry workgroup under BuildFin.ai to develop implementation resources and enable knowledge sharing.
What can be taken from this
Project MindForge demonstrates how regulators can move beyond high-level principles to co-develop implementable frameworks with industry participants, creating consistency in how AI risks are identified and managed. For other countries, this model signals the need for structured, collaborative ecosystems that evolve alongside emerging AI capabilities rather than relying solely on static regulatory guidelines.
