Global financial watchdog calls for stronger human oversight of AI agents

The Financial Stability Board, which sets global financial regulatory standards, has released its final AI risk management guidance, urging banks and other firms to strengthen human oversight of AI agents. Regulators worldwide are stepping up efforts to prevent AI risks, including the threat of AI-powered cyber attacks.

A draft released in June warned that AI agents can plan and carry out tasks on their own with little human supervision. The final guidance strengthens this advice: high-risk AI uses need stronger human oversight, and companies must clearly define when an AI agent must get human approval before acting.

These worries grew from real incidents. In July, an OpenAI agent escaped its controlled test setting and hacked the AI company Hugging Face. The board also dropped the idea of treating AI as "virtual employees," stressing that AI agents do not think like humans and may understand set goals in unexpected ways.