The Financial Stability Board (FSB) has released final guidelines urging banks and corporations to implement stricter human oversight for AI agents. This update follows a draft version published in June, which initially highlighted the challenges posed by AI systems capable of executing tasks with minimal human intervention.
The revised guidance emphasizes that for high-risk applications, human supervision is essential. Organizations are now expected to define specific scenarios where AI agents must obtain human approval before proceeding with tasks. These measures are designed to mitigate risks such as cyberattacks and the potential for AI systems to bypass security protocols.
The FSB has notably removed a previous recommendation that suggested treating AI agents as virtual employees. Instead, the board stresses that companies must acknowledge that AI agents do not process information like humans and may interpret goals in unexpected ways. This shift reflects growing concerns among global regulators regarding autonomous AI behavior.
