Artificial Intelligence and the Financial Centre: practical consequences
Read the PDF edition (2 pages)
For financial institutions in Singapore, the Guidelines of October 2026 turn artificial intelligence governance into a project with deadlines. Within twelve months they must have a framework for oversight, a complete inventory and a risk materiality methodology. Within twenty-four months they must apply lifecycle controls across all relevant use cases and ensure adequate capabilities. Institutions that belong to international groups may rely on group frameworks, provided these meet the Singapore expectations, and must be able to show how oversight of the Singapore operations is exercised.1
For clients and their advisers, the practical questions concern transparency and redress. A client who receives an adverse decision may ask whether artificial intelligence was involved, what data was used and how the decision can be reviewed. The FEAT Principles support such requests, although they do not create enforceable rights. For family offices and external asset managers, the main issue is the selection of providers: a portfolio management system or research platform with embedded artificial intelligence is third-party AI within the meaning of the Guidelines.
- Record every use of artificial intelligence, including AI embedded in purchased software, in an inventory with owner, purpose, data and provider.
- Assess each use case for impact, complexity and reliance, and document the result.
- Decide which use cases qualify for basic governance only and record the reasons.
- Assign each material use case to an identified senior manager under the individual accountability framework.
- Require human review for high-risk decisions affecting customers, such as credit, underwriting and advice, or use decision-deferral thresholds.
- Review contracts with providers for performance guarantees, data protection, audit rights and notification when artificial intelligence is introduced.
- Check that the use of customer data for training is covered by the PDPA exceptions or by consent, and that banking secrecy is respected.
- Prepare contingency plans and fallback options for high-risk use cases.
- Verify every authority, figure and citation produced by generative tools before it reaches a client, a regulator or a court.
- Train staff in proportion to their role and keep training materials current as the institution adopts more complex systems.
Notes
- MAS, Response to Feedback, 7 October 2026, paras 2.1, 4.3–4.4 and 13.5. ↩
For general information only. This article does not constitute legal or tax advice, does not replace advice on an individual case and creates no lawyer-client relationship. It reflects the legal position as of October 2026.