Singapore regulator calls for independent reviews of financial AI before use
Singapore’s Monetary Authority (MAS) has issued its first guidelines on AI risk management for financial institutions, calling for every AI use case to be independently reviewed before deployment. The move matters because MAS says AI’s complexity can make outcomes less predictable and harder to detect or correct, with generative and agentic AI posing further risks.
The guidance requires boards and senior managers to ensure AI is covered by risk frameworks, and calls for technology and cybersecurity checks, ongoing monitoring and current inventories of AI systems. Financial institutions remain accountable for third-party AI and should limit, suspend or replace services if their risks cannot be controlled. High-risk uses should have contingency plans, including manual alternatives; the guidelines take effect on 7 October 2027.
- MAS wants independent reviews of all financial-sector AI before deployment.
- Firms remain accountable for third-party AI services.
- The guidelines take effect on 7 October 2027.
New here? Start with this
Singapore's financial regulator has introduced guidelines requiring banks and investment firms to get independent assessments of their artificial intelligence systems before they start using them. The guidelines reflect concern that AI, particularly newer forms, can produce unpredictable results that may be hard to detect or correct once the systems are operating.
The rules require financial institutions to place AI oversight with their boards and senior leaders, carry out security checks and maintain records of all AI systems in use. Banks must treat AI from external providers the same way and be prepared to stop using those tools if the associated risks cannot be adequately managed.
Artificial intelligence is becoming widespread in financial services, where errors or failures can have serious consequences for people's savings and investments. Singapore's approach is among the first to specifically require independent review of AI before deployment in this sector, rather than relying on monitoring once systems are active.
Both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
MAS's requirements reflect necessary caution in critical infrastructure where AI failures can cascade through interconnected markets and destabilise the financial system. AI's inherent complexity makes outcomes difficult to predict and hard to correct post-deployment, so independent review before implementation helps catch risks before they harm consumers or markets. The stakes justify precaution: algorithmic bias, opaque trading decisions, or system failures don't just harm individuals but undermine confidence in the entire financial system, justifying strong regulatory oversight to ensure innovation proceeds responsibly.
The case against
Mandatory independent pre-deployment reviews create substantial compliance costs that disproportionately burden smaller institutions without clear evidence they prevent harms more effectively than existing market incentives and regulatory sanctions, which already push financial firms towards responsible AI deployment. Overly prescriptive rules may encourage compliance theatre rather than genuine risk ownership, and heavy regulation risks pushing innovation to less-regulated jurisdictions, diminishing Singapore's fintech competitiveness whilst slowing beneficial applications like AI-driven fraud detection. Principles-based regulation that defines safety outcomes rather than mandating specific review procedures would likely achieve oversight objectives more efficiently and flexibly.
Read the full article at the source →
Originally published by The Register as “Singapore’s central bank wants all FinTech AI use cases subject to independent review”.