AI in action · Regulation

Twelve Rules of Thumb for Banks Using AI: What the Financial Stability Board Finalised

On October 15, 2026 the Financial Stability Board published its final sound practices for adopting AI in finance. Human oversight should scale with how autonomous a system is, and dependence on a few AI providers is now a named risk.

Oct 16, 2026

The Financial Stability Board, which brings together finance ministries, central banks and regulators from 24 jurisdictions and reports to the G20, published its final report on the responsible adoption of artificial intelligence on October 15, 2026. It sets out 12 sound practices, split into two groups: governance across the whole organisation, and risk management through the life of each AI system.

The report is not a rulebook. It does not create legal obligations; instead, boards and senior management are "strongly encouraged to reference the sound practices", and national supervisors are expected to use them when they examine firms. That is how most international standards in finance start, and why they matter: what the FSB writes down tends to reappear, a year or two later, in national guidance.

Three of the practices stand out for markets. Practice 10 asks that human oversight be scaled to the autonomy and risk of each use case, a direct response to what the FSB calls the "rapid advances — particularly in generative AI (GenAI) and agentic AI". Practice 11 deals with cyber and technology risk specific to AI systems. Practice 12 covers dependence on outside AI providers, including concentration risk: if many banks rely on the same few models and cloud platforms, a single outage or change of terms becomes a problem for the whole system.

The others are less dramatic but just as practical: assessing how material each AI use is before deploying it, governing the data it is trained and run on, and being able to explain its outputs to customers and supervisors. The document went through a public consultation in June, with responses published in August, before the final version.

The FSB says the safe adoption of AI will remain on its agenda next year. For investors, the point is that the conversation among regulators has moved from whether AI creates risks to which controls they expect to see — and the firms that cannot show them will be the ones supervisors visit first.

Sources: Financial Stability Board — Sound practices for responsible adoption of AI: final report (Oct 15, 2026); Financial Stability Board — FSB Chair sets out FSB work on AI and stablecoins (Oct 2026)