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

Federal Reserve Governor Michael S. Barr has a question for every bank excited about AI agents: how do you know the agent did what the customer actually wanted?

Barr said banks will face challenges making sure agentic AI follows customer instructions. For an industry built on authorization, audit trails and liability, it touches nearly everything.

What Barr is worried about

Agentic AI refers to autonomous systems that can carry out multi-step tasks on their own. Think less chatbot, more digital assistant that can log in, move money and make decisions without checking back at every step.

A customer might tell an agent to “pay my bills.” The agent then has to decide which bills, from which account, in what order and by when. Every judgment call in that chain is a spot where intent and execution can drift apart.

In an address at the Singapore Fintech Festival on November 11, 2025, Barr said large banks were exploring generative AI for core work such as credit assessment, fraud detection and trading.

He also set a high bar for those deployments. Barr said outputs need to be controlled, precise, explainable and replicable.

Falling short of that standard creates real risk, according to the research summary of Barr’s remarks. It cited potential bias in lending decisions and unintended effects such as trading algorithms that could feed market volatility.

Banks are already pulling up the drawbridge

Between October 13 and 20, 2026, four major US banks announced restrictions on third-party AI agents logging into customer accounts.

JPMorgan Chase, Citigroup, Charles Schwab and Wells Fargo all moved to limit that access. The cited concerns were unauthorized transactions and murky liability for actions taken by so-called rogue agents.

Wells Fargo went a step further on the legal side. Under updated terms effective November 24, 2026, the bank shifted liability for unauthorized actions taken by AI agents onto customers.

As of late 2026, dozens of banks are building their own internal agentic tools while limiting outside agents.

A regulatory gap that banks are filling themselves

In April 2026, US interagency guidance explicitly excluded generative and agentic AI from model risk management requirements.

The Financial Stability Board has tried to frame the distinctive characteristics of agentic AI. Without direct rules, though, banks are handling the risks through internal governance and contract changes.

What this means

For banks, Barr’s warning lands as a signal about where supervisory attention is heading. A Fed governor publicly naming instruction-following as a challenge suggests examiners may eventually ask how banks verify that agents act within customer intent.

For fintech companies building consumer AI agents, the October 2026 restrictions from four of the biggest names in US finance cut off a direct route into customer accounts, at least for now.

For consumers, the stakes come down to accountability. The Wells Fargo terms show that, absent regulation, the default may be that customers absorb losses from agent errors.

Things to watch include whether the Fed or other agencies move to bring agentic AI into formal risk frameworks, and whether more banks follow Wells Fargo in shifting liability onto customers. How the Financial Stability Board’s work develops could also shape rules beyond the US.