Federal Reserve Governor Michael Barr said on Monday that AI agents acting for bank customers could speed up deposit runs to the point where they "unfold in minutes," and that major banks already report agents trying to take actions in retail accounts without the bank being able to confirm the customer approved them.
Barr made the remarks in prepared text for the Seventh Annual Association for Computing Machinery International Conference on Artificial Intelligence in Finance in Milan. The Fed published the speech the same day. He noted that the views were his own and not necessarily those of his colleagues on the Board.
Agents moving deposits on their own
The warning centers on speed. "Where customers once needed to execute transfers, AI agents could instantaneously detect yield differentials or stability signals and autonomously move funds at high speed, potentially accelerating runs that unfold in minutes, overwhelming traditional liquidity management and regulatory intervention timelines," Barr said.
He described the account activity as one of several new risks that "continue to surface." According to the speech, major banks "now report finding AI agents attempting to take actions in retail customer accounts, without the bank being able to verify authenticity, consent, and alignment with customer instructions." Barr did not name the banks or say how often this happens.
His proposed response is a set of guardrails for agentic finance. He called for strict authentication, explicit consent checks for specific actions, technical controls that limit what agents can do, and constraints based on both rules and values. He said new technology is needed to make sure agents follow established rules, which he called "especially critical as agentic commerce expands."
Trading herds and cyber risk
Barr also looked at AI in wholesale trading. "If traders rely on a small set of similar AI models trained on similar data, they face the risk of herding behavior," he said, with strategies that could "synchronize reactions to news and other information." A market full of similar models that share the same flaw "risks failures with systemic consequences," according to the speech.
He cited theoretical research showing that some reinforcement learning trading algorithms, known as Q-learning algorithms, can develop collusive behavior without explicit intent, and can learn to manipulate markets through tactics such as spoofing. Bad actors could also feed misleading market data or news to many similar AI trading systems at once, he said.
On security, Barr said incidents this year have shown AI models treating security boundaries "as obstacles to overcome rather than rules to be followed," which raises questions about how well rules will work even against agents told to follow them. He added that defenders are using AI to find and patch flaws faster, but "attackers only need to be right once, while defenders need to be right all the time."
A rising concern inside the Fed
The speech builds on the Fed's Financial Stability Report, released Friday. In the report's survey of market contacts, conducted in September and October, 87% cited artificial intelligence as a risk to U.S. financial stability, up from 50% in the spring survey, according to the report. AI was the most cited risk, with respondents pointing to high equity valuations and AI-enabled cyberattacks among their concerns.
Barr said on-balance-sheet debt at the seven largest AI firms is expected to exceed $800 billion before the end of the year, and that the report puts the potential maximum value of their off-balance-sheet obligations above $3 trillion.
Reuters reported that Barr's remarks came hours before Fed Chair Kevin Warsh was set to open a conference at the San Francisco Fed on AI's effects on jobs and the economy.
The comments land as consumer agents gain the ability to spend money. AgenticFeed reported last week on OpenAI's ChatGPT Wallet, which lets agents complete purchases for users, and card networks such as Mastercard have opened agent payment tools to developers. Barr said AI can only succeed in financial services over the long term with what he called "appropriate guardrails."
Source: Federal Reserve Board, “Speech by Governor Barr on some potential financial stability implications of artificial intelligence”
Additional sources: Federal Reserve Board, “Financial Stability Report, November 2026”; Bloomberg Law, “Fed's Barr Says AI Agents Could Fuel Future Bank Runs”; Reuters via Investing.com, “Fed's Barr warns on AI risks to financial system”
