Field report E-19 / L7 · sheet /macro
Boardroom MacroFiled from the territory
Fed's Barr warns AI agents could trigger bank runs in minutes
Fed Governor Michael Barr warned AI agents could autonomously move deposits at machine speed, causing runs in minutes. He cited $3 trillion in off-balance-sheet AI-linked liabilities and herding risks.
- Grid
- E-19 / L7
- Surveyed
- November 20, 2026
- Reading
- 4 min read
- Correspondent
- Sophie Lindqvist

Field notes
- Fed Governor Michael Barr said AI agents could accelerate bank runs that unfold in minutes.
- The Fed's Nov. 13 Financial Stability Report puts off-balance-sheet AI-linked liabilities above $3 trillion.
- On-balance-sheet debt of the seven largest AI firms is expected to exceed $800 billion by year's end.
- Barr flagged more than $100 billion in 'circular financing' by cloud and chip firms into AI labs.
- NatWest, Bank of America, ING and Capital One warned in September about AI shopping-agent risks.
Federal Reserve Governor Michael Barr warned on Monday that AI agents (autonomous software acting on a user's behalf) could compress bank runs into minutes, moving deposits faster than banks and regulators can respond. Speaking at a conference in Milan, he put a number on the stakes elsewhere in the AI buildout: more than $3 trillion in potential off-balance-sheet liabilities.
"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, according to Bloomberg. The result, he argued, would be runs "potentially accelerating runs that unfold in minutes, overwhelming traditional liquidity management and regulatory intervention timelines."
The claim is not hypothetical in his telling. Barr said major banks already report finding AI agents attempting actions in retail customer accounts — attempts the banks cannot verify for authenticity, consent, or alignment with the customer's actual instructions. He did not say, notably, whether any such attempt succeeded, nor how many banks reported the pattern. The absence of a figure is worth flagging: this is a warning built on early anecdotes, not a documented incident.
What does this change for banks and regulators?
Barr's answer is a set of safeguards, and it is more specific than most regulatory speeches on AI. He called for:
- tight authentication of agents acting on customer accounts;
- verification of explicit consent for particular actions, not blanket permission;
- technical limits on what agents can do, enforced in code rather than policy;
- restrictions grounded in both rules and values.
Whether banks can implement any of this before agent-driven transfers become common is the open question, and Barr did not address it. He described himself as an optimist about AI's long-term promise, with a condition attached: AI strengthens the financial system only if it also supports financial stability.
The $3 trillion liability question
The speech ranged beyond deposit flight to the financing behind the AI investment boom. Barr cited the Fed's latest Financial Stability Report, issued November 13, which estimates the potential maximum nominal value of off-balance-sheet liabilities tied to the AI buildout now exceeds $3 trillion. On the balance sheet, he said, the combined debt of the seven largest AI firms is expected to pass $800 billion before year's end.
He also pointed to more than $100 billion in cumulative equity investments by large cloud-computing firms and chipmakers in AI labs and cloud start-ups. Barr labeled this "circular financing" — the smaller firms use the invested money to rent or buy chips and computing power from the very investors backing them. Revenue flows back to the investor; the ecosystem books growth. What he did not quantify is how much of the sector's apparent expansion depends on that loop continuing.
Herding risk in AI-driven trading
On markets, Barr raised a different structural concern: traders relying on a small set of similar AI models face herding behavior, where strategies converge and react to news in sync. A market populated by many similar models sharing the same flaw, he said, risks failures with systemic consequences. He offered no estimate of how concentrated model use already is among trading desks.
The concern has precedent in non-AI form — correlated strategies were implicated in past flash events — but Barr's framing puts model homogeneity, not leverage, at the center.
The warnings are stacking up
Barr is not alone. A coalition of banks including NatWest, Bank of America, ING and Capital One warned in September that AI-powered shopping agents create new consumer risks around scams, fraud and data privacy. The banks said the technology is moving faster than industry standards and consumer protections — a frank admission that the industry's own defenses lag the deployment curve.
The International Monetary Fund added its own caution in October, warning that a crash in AI-related stocks could threaten global financial stability. The IMF cited rising debt levels, concentrated market exposures, and the same circular financing arrangements Barr flagged in Milan.
What links the three warnings is timing. The Fed, the IMF and the banks themselves are each describing the same gap — AI deployment speed outrunning the machinery of oversight — from different angles. Barr's contribution is the sharpest on mechanism: deposits, not stocks, are where machine-speed flight would hit first, and liquidity management is built for days, not minutes.
via qz.com (Source)
More from Sophie Lindqvist
Show full bio
Correspondent covering money, funding and market shifts around corporate strategy and chief-executive leadership at CEO Frontier.
515 articles
Adjacent sheets
- AI-Patched Software Is Likely Behind This Week's Outage Wave
- Ally, Chase and Citi Tell Customers: You Liable for AI Agent Errors
- Fed Prepares December Hike to Cool AI Boom It Admits It Cannot Reach
- Fed Finds $3 Trillion in Off-Balance-Sheet AI Commitments
- Fed Models Three AI Futures for Jobs, Growth and Rates