AI Agents Could Lead to a Deposit Run Occurring in Just Minutes, Fed Governor Warns

MILAN, Italy—Artificial intelligence agents acting on behalf of bank customers could move deposits so quickly during a crisis that bank runs unfold in minutes, potentially overwhelming financial institutions’ liquidity defenses and regulators’ ability to respond, Federal Reserve Gov. Michael Barr is warning.

In remarks prepared for an artificial intelligence finance conference in Milan, Barr said autonomous software could monitor changes in interest rates or signs of financial instability and transfer customer funds without waiting for individuals to initiate transactions.

Michael Barr

“Where customers once needed to execute transfers, AI agents could instantaneously detect yield differentials or stability signals and autonomously move funds at high speed,” Barr said, warning that the technology could accelerate deposit withdrawals beyond traditional response timelines.

Barr also said major banks have reported AI agents attempting to take actions in retail customer accounts without institutions being able to verify the agents’ authenticity, customers’ consent or whether the actions match customers’ instructions.

Call for Stronger Safeguards

He called for stronger safeguards, including strict authentication, verification of consent for specific transactions and technical limits on what AI agents are permitted to do.

According to Reuters, Barr’s remarks followed the Federal Reserve’s latest Financial Stability Report, in which 87% of surveyed contacts identified AI as a significant risk to financial stability. Their concerns included cyberattacks, increasingly complex financing for AI investments and the possibility that disappointing returns from AI could trigger broader stock-market losses.

Bloomberg also reported on Barr’s warning that autonomous agents could destabilize banks by shifting deposits in response to market signals or changes in yields.

Barr said AI’s risks extend beyond deposit flows. If financial firms depend on similar AI models trained on similar information, their trading decisions could become synchronized, amplifying market swings. He also warned that some self-learning algorithms could develop collusive or manipulative trading behavior.

The Cybersecurity Challenge

Cybersecurity poses another challenge, he said, because AI can help attackers find and exploit vulnerabilities even as financial institutions deploy the same technology to strengthen their defenses.

In consumer lending, Barr cautioned that AI models can be difficult to explain and may reproduce discrimination through biased data or indirect proxies for protected characteristics. Financial institutions must maintain effective model validation, fairness monitoring and consumer dispute processes, he said.

Barr emphasized that AI could improve financial services, expand access to credit and lower costs, but said those benefits depend on controls that preserve consumer protection and financial stability.

His prepared remarks did not address the outlook for interest rates or monetary policy, Reuters reported.

Facebook
Twitter
LinkedIn

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.