MAS Updates AI Governance Framework for Singapore FSI with Mandatory Explainability Requirements for Credit and AML AI

The Monetary Authority of Singapore has released an updated AI governance framework for financial institutions, introducing mandatory explainability requirements for AI systems used in credit decision-making, anti-money laundering (AML) surveillance, and trade surveillance — expanding the prescriptive elements of Singapore's FEAT (Fairness, Ethics, Accountability, and Transparency) principles from guidance to specific documentation and audit trail requirements.

The updated framework's explainability requirements affect three categories of FSI AI deployment: credit scoring and lending decisions (AI outputs affecting customer access to credit must be explainable to the affected customer and to MAS supervisors); AML transaction monitoring (AI flagging suspicious transactions must have auditable decision logic accessible to MAS examination); and algorithmic trading surveillance (AI identifying potential market manipulation must provide explainable reasoning trails). For Singapore FSI institutions deploying these AI systems, the update requires governance documentation review and, in some cases, model re-architecture to produce the required explanation outputs.

The practical compliance implication for APAC FSI teams is a need to evaluate their current credit, AML, and surveillance AI systems against the updated explainability standards before the compliance deadline. Black-box machine learning models that produce high-accuracy predictions without interpretable reasoning — a class that includes many legacy gradient boosting and deep learning credit scoring models — may require replacement with interpretable model architectures (logistic regression, decision trees, or XAI-augmented gradient boosting) or post-hoc explainability tooling that can generate MAS-acceptable explanation outputs.

MAS's updated framework also introduces vendor AI governance requirements: Singapore FSI institutions using third-party AI vendors for covered applications must ensure the vendor can provide the explainability documentation that MAS requires, creating a due diligence requirement for vendor selection and contract terms in Singapore FSI AI procurement.