MAS Financial AI Governance
The Monetary Authority of Singapore’s AI governance stack: FEAT Principles, Veritas Toolkit, MindForge, and AI Risk Management Guidelines.
Why MAS Leads Financial AI Governance
Singapore’s central bank and financial regulator has built the most mature sector-specific AI governance stack in Asia-Pacific.
Central Bank + Regulator
The Monetary Authority of Singapore serves as Singapore’s central bank and integrated financial regulator. MAS supervises banks, insurers, capital markets firms, and fintech companies. All financial institutions (FIs) operating in Singapore fall under MAS jurisdiction.
Progressive Evolution
MAS AI governance has evolved across four phases: voluntary principles (2018), industry assessment tools (2021-2023), formal supervisory guidelines (2025), and an operational risk management toolkit (2026). Each layer builds on the previous one without replacing it.
FEAT Principles (2018)
Fairness, Ethics, Accountability, and Transparency. The foundation that all subsequent MAS AI initiatives build on.
Fairness
AI should not result in unfair treatment of customers based on personal attributes. Financial institutions must monitor for and address discriminatory outcomes in AI-driven decisions like credit scoring, insurance pricing, and fraud detection.
Ethics
Use of AI should be aligned with the firm’s ethical standards and comply with applicable regulations. Ethical considerations should inform every stage of the AI lifecycle, from data collection through model deployment.
Accountability
Clear governance and internal oversight for AI use. Roles and responsibilities must be defined so that individuals and committees are accountable for AI outcomes, including escalation pathways and remediation processes.
Transparency
Appropriate disclosure to customers about AI use in decisions affecting them. Customers should understand when AI is being used, what data is involved, and how they can seek recourse if they believe a decision is incorrect.
Veritas Toolkit (2021-2023)
The first responsible AI assessment toolkit built specifically for the financial industry. Open source. MAS-led consortium.
Assessment Methodologies
Veritas provides structured assessment methodologies that financial institutions can use to evaluate their AI systems against the FEAT Principles. Version 2.0 (2023) includes methodologies for all four FEAT dimensions, enabling systematic validation of fairness, ethics, accountability, and transparency in financial AI. For organizations already using AI Verify, Veritas complements the national testing toolkit with financial sector-specific assessment criteria.
Open Source on GitHub
The Veritas toolkit is available as an open-source project (veritas-toolkit), making it accessible to any financial institution globally. This is the first responsible AI toolkit developed specifically for the financial sector, and the open-source model encourages adoption beyond Singapore’s borders.
AI Risk Management Guidelines (November 2025)
Consultation paper proposing formal supervisory expectations for AI use across the entire financial sector.
GenAI-Specific Considerations
The guidelines explicitly address generative AI and agentic AI risks: hallucination in customer-facing outputs, prompt injection attacks, data leakage through LLM interfaces, and the need for human oversight on high-impact GenAI decisions. FIs using GenAI face additional validation requirements.
Third-Party AI Accountability
Reliance on third-party AI vendors, cloud providers, or open-source models does not reduce a financial institution’s accountability. FIs remain fully responsible for AI outcomes regardless of whether the model was built in-house or procured externally. Vendor risk assessment is mandatory. Data handling by third-party AI vendors must also comply with PDPA obligations.
Project MindForge (2023-2026)
MAS-industry consortium of 24 leading financial institutions. Phase 2 concluded March 2026 with a published AI Risk Management Toolkit.
Operationalization Handbook
Detailed, step-by-step implementation guidance for AI risk management in financial institutions. Covers policy design, governance structures, risk assessment workflows, and monitoring procedures.
Executive Handbook
Strategic considerations for board members and senior management. Focuses on AI risk appetite, investment decisions, organizational readiness, and governance oversight responsibilities.
Implementation Examples
Real-world case studies from participating financial institutions. Shows how banks and insurers have implemented AI risk management controls across credit scoring, fraud detection, and customer service applications.
Coverage Scope
- Traditional AI and machine learning models
- Generative AI applications (LLMs, image generation)
- Agentic AI systems (autonomous decision-making agents)
- Third-party and vendor-supplied AI models
Consortium Members
24 leading banks, insurers, and capital markets firms participated in the MindForge consortium. The consortium approach ensures that the toolkit reflects actual operational challenges and real-world implementation patterns, not just theoretical best practices.
Phase 2 concluded in March 2026 with the publication of the complete AI Risk Management Toolkit.
MAS AI Governance Timeline
Eight years of financial sector AI governance, from voluntary principles to operational toolkits.
Related Tools
Practical tools to help financial institutions assess their AI governance against MAS expectations.
MAS FEAT Self-Assessment Template
Evaluate your financial AI systems against all four FEAT dimensions. Structured assessment with scoring guidance, gap identification, and remediation planning for each principle.
Singapore vs. EU vs. NIST Regulatory Mapping
Side-by-side control mapping across Singapore (MAS FEAT, Model Framework), EU AI Act, and NIST AI RMF. Identifies gaps and overlaps for multinational financial institutions.