AI in Finance Radar: September to October 2026

Global reference

• Period: Mid-September to early October 2026
• Coverage: Regulators, banks and finance functions
• Format: Monthly roundup

This month’s key moves

• Singapore, 7 October: MAS issued final Guidelines on AI Risk Management. They take effect on 7 October 2027 and apply to all financial institutions, with expectations scaled to risk. MAS will consult on agentic AI in 2027.
• European Union, 23 September: ESMA will make digital innovation, starting with AI and tokenisation, a union-wide supervisory priority from 2027. It flagged output bias and reliance on a few AI providers.
• United Kingdom: HMRC’s 2025-26 annual report says AI and analytics helped protect GBP 10 billion of tax revenue.
• United States, 16 September: the Conference of State Bank Supervisors published an optional AI supervisory framework with an examiner guide and a low, moderate and high risk-tiering worksheet.
• Banks, 22 September: six global banks, including Bank of America and Capital One, published principles for agentic commerce, warning of risks to transparency, privacy and consumer choice as AI agents shop and pay.
• Treasury: a TIS and EuroFinance study found 46% of treasury teams evaluating AI but only 4% with AI in production. Cash forecasting is seen as the best use case.

Still on the calendar

• 2 December 2026: end of the EU grace period for marking AI-generated content.
• 7 October 2027: MAS AI guidelines take effect.
• 2 December 2027: EU AI Act high-risk obligations, including credit scoring.

What to do now

• Banks and insurers with Singapore operations: start the AI inventory now; a year is short.
• Treasury and payments teams: plan controls for payments started by AI agents.
• Everyone: review where AI vendors are concentrated in your stack.

Our take

The common theme this month is supervision moving from principles to inspection. Expect examiners to ask for your AI inventory before they ask about anything else.