MAS FEAT, AI Governance for Singapore Finance

MAS FEAT is a set of four principles, Fairness, Ethics, Accountability, and Transparency, published by the Monetary Authority of Singapore in 2018 to guide the responsible use of artificial intelligence and data analytics in the financial sector. While principles-based, FEAT is the reference framework MAS uses during supervisory review of AI deployments by licensed financial institutions.

• Fairness: AI-driven decisions are justifiable. Data and models are reviewed for bias. Individuals are not systematically disadvantaged.
• Ethics: AI usage aligns with the firm's ethical standards and broader societal norms. Use-case fit is examined before deployment.
• Accountability: clear internal ownership of AI decisions. Mechanisms exist for affected parties to challenge outcomes and for FIs to remediate.
• Transparency: AI usage is disclosed to affected customers. The decision rationale is explainable at an appropriate level for the audience.

FEAT principles alone are abstract. The Veritas Initiative is MAS’s industry consortium that operationalizes FEAT into concrete assessment methodologies. Veritas publishes open-source toolkits and case studies for specific use cases, such as credit risk scoring, customer marketing, fraud detection, and claims processing. A financial institution demonstrating FEAT alignment increasingly does so via a Veritas-style assessment rather than a narrative compliance memo.

All MAS-supervised financial institutions deploying AI in customer-impacting decisions: banks, insurers, capital markets services, payment service providers, financing companies, and licensed fintechs. Vendors and third-party service providers supporting AI deployments are expected to support FEAT compliance through documentation, audit logs, and Veritas-style assessment cooperation.

The Personal Data Protection Act (PDPA) governs how personal data is collected, used, and disclosed in Singapore generally. MAS FEAT governs how AI and data-analytics decisions are made in the financial sector specifically. They overlap on transparency and accountability, but FEAT is sector-scoped and broader: it covers AI fairness and ethics beyond PDPA’s data-handling concerns. Both apply to Nova Solutions deployments in Singapore.

• Black-box models without explainability, which fails Transparency. Any model used in a customer-impacting decision needs decision-rationale traces.
• Training data with unaddressed bias, which fails Fairness. Bias testing on age, gender, nationality, and income is required for credit and underwriting models.
• No clear ownership for outputs, which fails Accountability. Who at the FI signs off on the AI's decision logic, and who handles customer challenges?
• Hidden AI usage, which fails Transparency. Customers in scope must be aware AI is part of the decision and have a path to human review.
• No audit trail, which fails all four. Every customer-impacting AI interaction needs a permanent, timestamped, queryable record.

Nova for Finance is engineered around FEAT from the data layer up:

• Full interaction logs: every AI interaction permanently recorded with transcript, timestamp, and decision rationale. Transparency and Accountability.
• Explainability traces: model outputs include the factors that drove each decision. Transparency.
• Bias monitoring: demographic-bias checks run continuously on training data and live predictions. Fairness.
• Disclosure-by-default: voice agents identify as automated, and SMS templates include AI-disclosure where required. Transparency.
• Veritas-ready evidence: the dashboard surfaces all of the above in formats Veritas assessments expect. Ethics and Accountability.

Singapore-licensed FIs deploying Nova for Finance do not have to retrofit FEAT compliance. It is the baseline architecture.