Overview
- The Monetary Authority of Singapore published a 30‑page guidance document on Wednesday, October 7, 2026, that sets four core expectations for banks, insurers, asset managers and fintechs.
- Firms are explicitly held accountable for AI outcomes even when models or services come from third parties, and must obtain sufficient assurance or apply compensating controls or stop using the service if risks exceed their appetite.
- Institutions must identify AI uses, keep an inventory of systems and models, and assess risks at both the enterprise level and for each use case before applying controls.
- The guidance requires lifecycle controls such as data governance, model testing, human oversight, cybersecurity, monitoring and change management and sets phased deadlines that start on Oct 7, 2027 with full compliance due by Oct 7, 2028.
- MAS signalled further work on increasingly autonomous or agentic AI with a planned sector consultation in 2027, a step that could force vendors to provide stronger technical assurances and change how customers are protected.