Michael Barr, governor of the US Federal Reserve (Fed), is concerned about the consequences of advances in artificial intelligence (AI). In his view, autonomous AI agents could accelerate future bank panics by moving deposits at a speed never seen before.
Bank runs that could unfold in just a few minutes
At an international conference on artificial intelligence held in Milan, Italy, on November 16, 2026, Michael Barr, a member of the Board of Governors of the Federal Reserve, warned that autonomous AI agents could dangerously accelerate bank runs, meaning sudden, massive bank withdrawals.
In his speech, Governor Barr explains that these AI tools could instantly detect changes in yields or stability alerts. Whereas customers previously had to act manually to withdraw their deposits, autonomous agents could accelerate these withdrawals to the point that they "unfold within minutes, outpacing traditional liquidity management and regulatory intervention timelines."
Speech by Governor Barr on some potential financial stability implications of artificial intelligence: https://t.co/5WJ28iremn
Watch live: https://t.co/IIsyOn6Dkd
Learn more about Governor Barr: https://t.co/AZsNWhl2ze
- Federal Reserve (@federalreserve) November 16, 2026
In other words, an AI-accelerated bank panic could drain a bank before the regulator has time to react. The speed of digital transactions already complicates this task. The collapse of Silicon Valley Bank (SVB) on March 9, 2023 is a case in point: customers tried to withdraw 42 billion dollars in a single day, nearly a quarter of the bank's total deposits, leading to its placement under regulatory receivership the very next day.
This bank run, already considered the fastest in modern US banking history, nonetheless unfolded over several hours, through manual orders placed from mobile apps. Autonomous AI agents could concentrate the same movement into a few minutes, well before any regulatory intervention.
AI agents already active in bank accounts
The Fed governor also stresses that the risk is no longer merely theoretical. Since last month, it has been known that several major US banks have detected AI agents attempting transactions on individual customers' accounts without their explicit consent.
At least five major banks, including JPMorgan Chase, Citigroup, Bank of America, Wells Fargo and Morgan Stanley, now block third-party AI agents' access to their customers' accounts.
For Barr, the response consists of putting in place stricter authentication protocols, explicitly verifying consent and introducing technical safeguards that limit agents' capabilities. AI-driven decisions must remain "well controlled, precise in quantitative and legal terms, explainable and reproducible."
The systemic risk linked to AI goes beyond bank deposits alone
Beyond the risk of a bank run, Michael Barr points to broader systemic vulnerabilities linked to the expansion of AI infrastructure. According to the Fed's latest Financial Stability Report, on-balance-sheet debt of the seven largest AI companies is expected to exceed 800 billion dollars by year-end. Off-balance-sheet commitments, including long-term data center leases and energy supply contracts, now exceed 3000 billion dollars.
The governor also points to a phenomenon of "circular financing": major chipmakers such as Nvidia and hyperscalers, the cloud giants such as Amazon and Microsoft, inject more than 100 billion dollars into small start-ups. These firms reuse the funds to buy computing capacity. This mechanism could amplify a market shock if returns on AI investments were to disappoint.
Michael Barr also flags asymmetric risks in cybersecurity. Autonomous models would tend to treat safeguards as "obstacles to overcome rather than rules to follow." A problem already observed in the crypto ecosystem, where AI agents have managed to circumvent their own spending limits.
The Fed governor's warnings amount to saying that the adoption of agentic AI in finance is advancing faster than the establishment of safeguards. Regulators need to catch up before the next banking shock reveals the full extent of the problem.

