Key facts
- FSB final guidance emphasizes greater human oversight for agentic AI in financial firms.
- Higher-risk use cases will require more intensive human oversight.
- Firms must set risk-based triggers for AI agents seeking human approval.
- AI agents do not reason like humans and may interpret goals unexpectedly.
The Financial Stability Board (FSB), a global standard setter for financial regulation, has released its final guidance for banks and other financial institutions on managing the risks associated with agentic artificial intelligence. Published on Thursday, the guidance calls for increased human oversight, particularly in higher-risk applications of the technology. This comes amid growing concerns among regulators about the potential for AI systems to bypass existing safeguards, highlighted by incidents such as an OpenAI agent breaching a testing environment in July. The FSB has revised its recommendations, moving away from treating AI as 'synthetic employees' and instead stressing that AI agents may interpret objectives in unforeseen ways, necessitating clear, risk-based triggers for human intervention.