• The FSB finalised 12 sound practices for responsible AI adoption on 15 October 2026, after a June 2026 consultation.
• The practices fall into two groups: organisation-wide AI governance, and risk management across the AI lifecycle from development to deployment.
• The FSB describes the practices as a practical toolkit institutions can tailor, and strongly encourages boards and senior management to reference them.
• The FSB's own wording points to guidance rather than binding regulation, so any enforcement would depend on national authorities.
• The FSB says it will monitor developments and update the practices as needed, so the list may change.

The Financial Stability Board on 15 October 2026 finalised its sound practices for the responsible adoption of AI. The report follows a June 2026 consultation and sets out 12 practices. The FSB strongly encourages boards and senior management to reference them, but describes them as a toolkit institutions can tailor. How much that changes bank behaviour depends on what national supervisors do next.

Two groups of practices

The FSB's announcement sets out 12 practices in two groups. One group addresses organisation-wide AI governance. The other addresses risk management across the AI lifecycle, from development to deployment. The split means an institution is expected to think about AI at two levels at once: who is accountable across the whole organisation, and how each system is controlled over its life. A bank that could evidence only one level would have covered half of the structure the FSB describes. The announcement does not rank the practices by importance.

A toolkit, not a rulebook

The FSB describes the 12 practices as a practical toolkit that institutions can tailor, and it strongly encourages boards and senior management to reference them when developing their AI governance, risk management and control frameworks. On our reading, that wording points to guidance rather than binding regulation, and the PYMNTS headline about company-wide AI regulation reads more strongly than the FSB's own language. Any enforcement would have to come from national authorities choosing to adopt or cite the practices.

What the chair's letter adds

The FSB chair's October 2026 letter to G20 finance ministers and central bank governors lists AI among the risks the FSB is tracking. The letter page, as reviewed, does not set out a specific AI work programme or explain why the practices were issued. The FSB's announcement says it will monitor developments and update the practices as needed. Taken together, the sources support a narrow reading: AI is on the FSB's radar, and the practices are expected to be revisited rather than treated as final.

What banks should be able to show

This is The Ledger's own reading, not an FSB statement: the practical test for a bank is whether it can show, for each AI use case, who is accountable and how the system is controlled across its lifecycle. The FSB sources do not say how any specific institution, including neobanks, currently performs against the practices, so no comparison is drawn here. Readers should watch whether national supervisors cite the practices in their own expectations, and whether the FSB's updates change the list of 12.