
Fed's Barr Warns AI Agents Could Trigger Bank Runs in Minutes
Fed Governor Michael Barr says AI agents could move deposits "at high speed," and seven AI firms carry $800B on-balance-sheet debt with $3T in off-balance-sheet obligations.
- By
- Rebecca Stone
- Filed
- Channel
- AI & Compute
- Read
- 3 min read
Federal Reserve Governor Michael Barr said on Nov. 16 that AI agents acting for bank customers could make deposit runs faster than banks and regulators can handle — with on-balance-sheet debt at the seven largest AI firms expected to top $800 billion this year and their off-balance-sheet obligations exceeding $3 trillion.
Barr laid out the risk in prepared remarks at the Association for Computing Machinery's International Conference on AI in Finance in Milan. He described a scenario in which autonomous software, not panicked customers, drains deposits.
"Where customers once needed to execute transfers, AI agents could instantaneously detect yield differentials or stability signals and autonomously move funds at high speed, potentially accelerating runs that unfold in minutes, overwhelming traditional liquidity management and regulatory intervention timelines," Barr said.
Minutes, not days. That compression is the core of his warning.
Are banks already seeing this?
Yes, at least at the probing stage. Major banks "now report finding AI agents attempting to take actions in retail customer accounts," according to Barr — attempts banks cannot verify for authenticity, consent, or whether the action matches the customer's actual instructions.
His prescription: strict authentication, explicit consent checks for specific actions, and hard technical limits on what agents can do in customer accounts.
The speech did not address interest rates or the economic outlook. Fed Chair Kevin Warsh was scheduled to speak the same day at noon ET.
What worries him about AI trading?
Herding. "If traders rely on a small set of similar AI models trained on similar data, they face the risk of herding behavior," Barr said, with strategies that "synchronize reactions to news." A market full of similar models sharing the same flaw "risks failures with systemic consequences," he warned.
Barr also flagged:
- Bad actors could feed false trades, quotes or news to mislead many AI trading systems at once.
- Research shows some self-learning trading algorithms can learn to collude without explicit intent, or to manipulate markets through spoofing.
- The most expensive AI systems could hand the largest trading firms an edge and concentrate market power.
For now, Barr said his understanding is that generative AI "is not yet being used for autonomous trading at large scale." Still, "some caution is warranted" while the technology's development "may be galloping ahead of our understanding," he said. Regulators need to prepare before the gap closes.
How much debt backs the AI buildout?
Barr drew on the Fed's Financial Stability Report, released Friday, Nov. 13, which named AI the most-cited risk among surveyed market contacts. His figures, on the record:
- On-balance-sheet debt at the seven largest AI firms has grown sharply in 2026 and should exceed $800 billion by year-end.
- The potential maximum value of their off-balance-sheet obligations — long-term contracts, leases, residual value guarantees — "now exceeds $3 trillion."
- Hyperscalers and GPU makers hold more than $100 billion of equity stakes in AI labs and cloud start-ups that in turn rent or buy their chips and compute. Barr called this "circular financing."
"Given the still-limited amount of debt backing the current boom, I'm less concerned about direct losses," Barr said. The sharper danger sits elsewhere: with hedge fund leverage and risk appetite "apparently quite high," a sharp reversal "could lead to rapid deleveraging and, potentially, self-reinforcing fire sales." Dealers' prime brokerage exposures to firms with large AI positions need monitoring, he said.
One hardware datapoint offered a sliver of comfort: lease rates on older chips "appear to be holding up" for now.
Who else is sounding the alarm?
Barr is not alone this month. ECB Governing Council member Boris Vujcic named AI among the leading risks to euro zone stability, and IMF chief Kristalina Georgieva pointed to more than $1 trillion of AI circular financing.
Barr noted the Milan speech was his fourth in the past year devoted entirely to AI — a signal of how firmly the technology has moved onto the Fed's financial-stability agenda, and a marker that supervisory scrutiny of AI-driven flows and buildout financing will likely intensify rather than fade.
Source: mlq.ai
More from Rebecca Stone
Show full bio
Correspondent covering technology and product news in Semiconductors, chip manufacturing, hardware, AI compute at Chip Dispatch.
1,037 articles
Related articles
u-s-banks-restrict-ai-agents-after-account-access-attempts
U.S. banks restrict AI agents after account access attempts
fed-chair-warsh-to-grill-ai-s-economic-claims-in-san-francisco
Fed Chair Warsh to Grill AI's Economic Claims in San Francisco
87-of-fed-survey-contacts-now-call-ai-a-top-financial-risk
87% of Fed Survey Contacts Now Call AI a Top Financial Risk
ai-firms-borrow-hundreds-of-billions-for-chips-as-wall-street-bankrolls-buildout-b0d37b4e
AI Firms Borrow Hundreds of Billions for Chips as Wall Street Bankrolls Buildout



