Federal Reserve Governor Michael Barr has warned that artificial intelligence agents could dramatically increase the speed of bank runs, allowing deposit withdrawals to unfold within minutes and potentially overwhelming traditional protections in the banking sector.
Major banks face risks from AI-initiated transactions
Speaking to financial industry leaders at an AI finance conference in Milan on November 16, Barr said that large banks are already encountering cases where AI agents attempt to carry out actions in customer accounts. According to Barr, these institutions cannot always verify whether the software initiating such actions truly represents the account holder or possesses the necessary authorization.
The Federal Reserve governor did not disclose the names of the institutions involved, nor did he specify how often these incidents were occurring. He did not confirm whether the attempted actions led to unauthorized transfers. Nonetheless, Barr emphasized that the presence of such incidents reflects new challenges for the sector as AI-powered banking tools proliferate.
Many major financial institutions are now experimenting with autonomous AI agents that process transactions, manage account details, and make financial decisions, often requiring minimal or no human intervention.
Barr highlighted that the core issue lies in determining whether a software agent genuinely has the authority to access or manage a customer’s funds, especially as these systems become more sophisticated and capable of interpreting and executing complex instructions on behalf of clients.
Unlike conventional bank transaction systems, these AI agents can make decisions, initiate transfers, and connect with a range of other financial services, creating broader opportunities for both innovation and potential misuse.
Barr warned that banks could struggle to keep pace with AI-driven attempts to access accounts, particularly as software grows more sophisticated and can make rapid, complex financial decisions independently.
Mini dictionary: Federal Reserve — The central banking system of the United States, directing monetary policy, overseeing financial institutions, and helping maintain financial stability.
AI agents could accelerate and amplify systemic risks
Barr also raised concerns about how AI agents could fundamentally change the dynamics of a bank run. Traditionally, customers would have to manually assess risk, decide to move their deposits, and initiate withdrawals themselves. The adoption of autonomous AI could allow these steps to be conducted automatically, with software drawing its own conclusions from market or institutional signals and acting before most customers are even aware.
If many banking agents respond simultaneously to distress signals, large outflows of deposits may quickly outpace a bank’s ability to respond with risk mitigation measures, intensifying liquidity pressures. Even correctly authorized AI-powered withdrawals, executed across thousands or millions of accounts at once, could trigger unprecedented systemic risks.
Barr noted that even in currently existing payment systems where transactions are handled by agentic software with appropriate permissions, rapid and synchronized movements of funds at scale present new hazards.
If AI models across the financial sector react in similar ways to new market developments, automated tool-driven trading and withdrawal activity could amplify instability beyond past experiences.
| Traditional banking | AI-enabled banking |
|---|---|
| Manual withdrawals by customers, usually after news or risk assessment | Automated withdrawals by AI agents responding to signals in real-time |
| Withdrawal process takes hours or days | Potential for withdrawals to occur within minutes |
| Bank staff verifies large movements | Software makes decisions and executes actions autonomously |
The growing adoption of these technologies requires institutions to consider both security and stability implications more carefully.
Calls for enhanced safeguards and oversight
Barr advised banks to strengthen their authentication methods, require explicit customer consent for critical actions, and enforce technical restrictions on AI agent capabilities. He signaled that while AI may offer operational benefits, these advantages cannot come at the expense of security or financial system stability.
As banks evaluate the costs and risks associated with integrating AI-based systems, recent industry analyses have questioned whether the promised efficiency gains justify the increased complexity and exposure to novel threats.
Barr also referenced the Federal Reserve’s most recent financial stability assessment, which highlights risks related to the rapid expansion in the use of AI across financial markets and infrastructure.






