Bank regulators on both sides of the Atlantic are asking lenders hard questions about their AI agents.
Reuters reports, citing six people familiar with the matter, that the Federal Reserve and two other US bank regulators have begun questioning banks about how they deploy AI agents and what could go wrong. European regulators are running a parallel effort. The shared worry: a cyberattack or a rogue AI agent could trigger problems that ripple across the financial system. The source report does not specify which banks have been contacted or what, if any, new rules might follow.
Banks have spent the last two years racing to deploy AI agents for things like fraud detection, trading, and customer service, often faster than their risk teams can audit them. Regulators asking questions now, rather than after an incident, suggests they see agentic AI as a systemic risk category, not just a vendor due-diligence checkbox.
It's the same instinct that drove scrutiny of high-frequency trading algorithms after they caused flash crashes - except this time, the algorithms can write their own code.