Fed's Barr Warns AI Agents Could Spark Bank Runs in Minutes
Federal Reserve Governor Michael Barr says AI agents are already testing retail bank accounts, and autonomous transfers could drain them faster than regulators can react. In a Milan speech, he pushed for authentication, consent checks and hard technical limits on what agents are allowed to do.
- 2 sources report it: thedefiant.io, cryptopolitan.com.
- The source published it on Nov 16, 2026 at 10:18 AM.
Nothing open. Every check passed.
Barr delivered the warning in a Nov. 16 speech at the Association for Computing Machinery's seventh annual International Conference on Artificial Intelligence in Finance in Milan. Major banks, he said, are already reporting AI agents attempting actions in retail accounts while the bank cannot verify authenticity, consent, or whether the move matched the customer's instructions. The speech was titled 'Some Potential Financial Stability Implications of Artificial Intelligence.' He stressed he was speaking for himself, not for his colleagues on the Federal Reserve Board or its rate-setting committee, and announced no new rule or examination program.
His fix: strict authentication, consent checks for specific actions, and technical limits on agents. Without them, he warned, autonomous transfers triggered by higher yields or signs of stress could turn a bank run into a minutes-long event, overwhelming banks' liquidity management and regulators' response times. Barr said generative AI was not yet being used for autonomous trading at large scale to his understanding, but flagged a different danger — widespread reliance on similar models trained on similar data could synchronize trades, turn common flaws into systemic failures, and let manipulated news or quotes mislead many systems at once.
Cost is its own risk. Barr said the expense of advanced systems could concentrate market power among large firms even as AI tools broaden access to trading technology. 'Many of the most important questions about how AI will affect financial markets will be answered not in the abstract, but in the technical choices made about model design, training data, testing, and deployment,' he said.
Cybersecurity cuts both ways too, he said: AI helps defenders find and patch vulnerabilities, while also letting attackers exploit shared service providers and weaknesses that were previously too difficult to hit. For lenders, Barr urged explainable AI, continuous fairness monitoring, detailed model validation records and accessible dispute mechanisms. Alternative data could widen access to credit, he said, but could also act as a proxy for protected characteristics such as race or national origin.
Barr largely repeated the financing concerns from the Fed's November Financial Stability Report, which traced how leases, guarantees and circular financing link large technology companies to riskier borrowers. He cited its estimate of more than $3 trillion in potential maximum nominal obligations under long-term contracts, leases and guarantees — arrangements that can hide where risk sits and worsen selloffs when sentiment shifts. Direct losses worry him less, given the still-limited debt backing the boom; his concern is that a sharp reversal forces leveraged hedge funds to sell, creating self-reinforcing fire sales. Big banks' high capital levels offer protection, he said, but their exposures to counterparties holding large AI positions warrant monitoring.
TopicsMichael Barr · Federal Reserve · Federal Reserve Board · Association for Computing Machinery · Financial Stability Report
Written by AI from verified reporting.

