Financial Stability Board urges agentic AI safeguards amid 52% adoption
The Financial Stability Board (FSB) has urged banks and financial firms to consider safeguards and clearer boundaries for agentic AI (think smart systems that can make their own decisions).
The FSB is worried that these AI systems could go rogue, leading to data leaks or unauthorized actions.
Currently, 52% of financial-sector respondents in a Cambridge Centre for Alternative Finance survey reported active agentic adoption for things like spotting fraud or helping customers.
FSB draft recommendations open for comment
The big issue? These AI systems can make decisions way faster than humans can keep up with, making it tough to step in if something goes wrong.
The FSB suggests setting clear limits on what AI can do, requiring human approval for risky activities like big money moves, and even treating AI agents a bit like digital employees with their own guidelines.
Their recommendations aren't law yet (they are open for public feedback until July 22), but it shows regulators are taking AI risks seriously as the technology keeps evolving rapidly.