Federal Reserve’s Barr warns of agentic AI challenges for banks

Federal Reserve Governor Michael S. Barr has raised concerns about banks' ability to ensure autonomous AI agents follow customer instructions accurately. Speaking at the Singapore Fintech Festival on November 11, 2025, Barr highlighted that for an industry built on authorization, audit trails and liability, the challenges of agentic AI touch nearly every aspect of banking operations.

Agentic AI challenges

Agentic AI refers to autonomous systems that can carry out multi-step tasks independently, going beyond simple chatbots to digital assistants that can log in, move money and make decisions without constant human oversight. Barr emphasized that when a customer instructs an agent to 'pay my bills,' the system must make numerous judgment calls about which bills to pay, from which account, and in what order—creating potential gaps between intent and execution.

Bank restrictions and liability shifts

Between October 13-20, 2026, four major US banks—JPMorgan Chase, Citigroup, Charles Schwab and Wells Fargo—announced restrictions on third-party AI agents accessing customer accounts. Citing concerns about unauthorized transactions and unclear liability for actions by 'rogue agents,' these moves represent a defensive stance by financial institutions.

Wells Fargo went further by updating its terms effective November 24, 2026, to shift liability for unauthorized AI agent actions onto customers. This approach highlights how banks are addressing regulatory gaps through internal governance changes rather than waiting for formal oversight.

Regulatory landscape

In April 2026, US interagency guidance explicitly excluded generative and agentic AI from model risk management requirements. Without direct regulations, banks are handling risks through internal policies and contract modifications. The Financial Stability Board has attempted to define agentic AI's distinctive characteristics, but formal rules remain in development.

Barr's warning signals where supervisory attention may be heading. A Fed governor publicly identifying instruction-following as a challenge suggests examiners may eventually require banks to demonstrate how they verify AI agents act within customer intent. The industry will be watching whether regulatory frameworks evolve to address these unique challenges.