The Financial Stability Board finalises 12 sound practices for financial institutions' use of AI, with human oversight scaled to each use's risk and autonomy
The G20's financial watchdog strongly encourages bank boards to reference them; they also cover cyber risk and dependence on outside AI providers.
The Financial Stability Board, which coordinates financial regulators for the G20, published its final report on sound practices for the responsible adoption of AI on 15 October. Its release says: “The report identifies 12 sound practices”. It reflects feedback on a consultative version published in June 2026.
The practices come in two groups. The first covers governance across the organisation: board oversight, clear roles, AI in risk frameworks and continuous learning. The second covers the life of each AI system, from assessing how material a use is through data governance, explainability and performance testing.
Sound practice 10 asks institutions for human oversight relevant to the materiality, risk, autonomy, complexity and explainability of different AI use cases. Practice 11 covers AI-related cyber and ICT risks, and practice 12 covers risks from third-party AI dependencies, such as reliance on a handful of model and cloud providers.
The release says the FSB “strongly encourages the board and senior management of financial institutions to reference the sound practices”. It calls them a flexible toolkit, and says it will update them as the technology develops.
Some press headlines framed the report as a call for human oversight of agentic AI. The report names agentic AI among recent advances, but its oversight practice applies to every AI use, scaled to its risk. It follows Singapore's central bank, which finalised its own AI risk guidelines for financial institutions on 8 October.
- Confirmed The FSB's final report, published 15 October, identifies 12 sound practices and reflects feedback on a June 2026 consultative version. Financial Stability Board
- Confirmed Sound practice 10 calls for human oversight relevant to the materiality, risk, autonomy, complexity and explainability of each AI use case. Financial Stability Board
- Confirmed Practices 11 and 12 cover AI-related cyber and ICT risks and risks from third-party AI dependencies. Financial Stability Board
Courts & policyMoney & markets
Today in the October 16, 2026 edition · front page