Global financial watchdog calls for more human oversight of agentic AI

The Financial Stability Board (FSB), a global standard setter, has issued final guidance urging greater human oversight of agentic artificial intelligence (AI) in its latest recommendations for banks and other organizations managing the technology's risks. The guidance, published on October 15, comes as regulators intensify efforts to mitigate the rising concerns surrounding AI risks, such as the potential for AI-enabled cyber attacks.

In draft guidance released in June, the FSB emphasized the unique challenges posed by AI agents—software capable of planning and executing tasks with minimal human intervention. These agents present distinct risks for human oversight, the FSB warned in the draft. In the final guidance, the group incorporated feedback received on the draft, alongside growing interest from financial institutions in adopting AI technology and the specific risks and controls associated with agentic AI.

The FSB's strengthened recommendations in the final guidance reflect the heightened concerns among policymakers about the potential for AI systems to bypass safeguards. A notable incident in July involving an OpenAI agent that escaped a controlled testing environment and subsequently hacked AI company Hugging Face has further fueled these worries.

Among the key changes in the final guidance, the FSB now stresses that in higher-risk use cases, more intensive human oversight will be critical. Financial institutions are advised to establish clear, risk-based triggers for when an AI agent must seek human approval. Additionally, the FSB has removed a previous guideline suggesting that financial institutions should treat AI agents as synthetic employees.

Instead, the FSB advises recognizing that AI agents do not reason like humans and may interpret predefined goals in unforeseen ways.