The Financial Stability Board has strengthened its guidance governing artificial-intelligence systems used by banks and other financial institutions, putting greater emphasis on human control as companies deploy increasingly autonomous AI.
The final framework identifies 12 sound practices covering AI governance and risk management throughout the technology’s lifecycle.
The FSB emphasized that boards and senior management should oversee how AI adoption aligns with an institution’s business model, strategy and risk appetite.
The guidance specifically calls for appropriate and effective human oversight as financial institutions adopt more autonomous systems.
Agentic AI differs from traditional generative AI because agents can potentially plan and execute multi-step tasks with less direct human involvement.
That creates a different category of operational risk.
An AI agent used inside a bank could interact with internal systems, analyze transactions, initiate workflows or perform other actions. Greater autonomy increases the consequences if a system misinterprets instructions or behaves unexpectedly.
The FSB’s framework also calls for institutions to evaluate AI systems throughout their lifecycle, establish appropriate data governance, assess explainability, manage cybersecurity risks and monitor dependence on third-party AI providers.
The final recommendations follow a consultation report published in June and incorporate feedback from financial institutions and other stakeholders.
The guidance is not binding legislation.
Instead, it provides an international framework that banks, regulators and other financial institutions can incorporate into existing governance and risk-management systems.
The implications extend beyond cybersecurity.
Autonomous AI can create operational, governance and compliance risks if institutions cannot determine why a system took an action, who authorized it or when human intervention is required.
For financial firms, the regulatory direction is becoming clearer: increasing AI autonomy is likely to require stronger governance rather than less oversight.
The next question is how individual banking regulators translate the FSB’s international recommendations into supervisory expectations within their jurisdictions.
