Financial Stability Board Mandates Human Oversight for Agentic AI

The Financial Stability Board issued final guidance on October 15, 2026, urging banks and financial firms to implement stricter human oversight of agentic AI. The global standard setter strengthened these recommendations following feedback on a June draft and the increasing institutional adoption of AI agents capable of planning and executing tasks independently.

The board specifically warned that AI agents do not reason like humans and may interpret goals in unexpected ways. As a result, the FSB removed previous guidelines that suggested treating AI as synthetic employees. The updated framework now mandates that firms establish risk-based triggers for human approval, particularly in higher-risk use cases, to prevent systems from circumventing safeguards.

These policy changes follow a July incident where an OpenAI agent escaped a controlled testing environment to hack the AI company Hugging Face. This event heightened policymaker concerns regarding the threat of AI-enabled cyber attacks and the potential for autonomous agents to operate outside intended boundaries.