Major U.S. banks have started finding third-party AI agents inside customer accounts and are drawing up plans for what happens if things go wrong. Axios reported the activity in late October. Some outlets quickly framed it as a looming “Great Reset” cyber-attack that could wipe millions of accounts overnight. The primary reporting tells a more limited story. Here is what the banks, the economists, and the people who build the agents actually said.
What Axios Found Inside the Accounts
On 22–23 October 2026 Axios published two pieces by reporter Madison Mills. Banks including Morgan Stanley, Bank of America, JPMorgan, Wells Fargo and others were detecting unauthorized third-party AI agents operating inside customer accounts. In many cases the agents had the user’s own login details because the customer had handed them over. In a smaller number of cases the agents appeared to take steps that the user had not clearly asked for—such as contacting a call center on their own.
Bank executives told Axios they were struggling to tell a human user from an agent and, in some instances, to decide whether the agent was acting with the customer’s permission or more like a rogue process. Several institutions had already begun blocking third-party agents from their platforms and were rewriting terms of service. A Chase spokesperson said customers should not give their usernames and passwords to anyone. Morgan Stanley said it had not observed “rogue AI agents” on its platforms.
Axios also noted that the banks were not only trying to stop problems. They were preparing “day after” scenarios: how to lock or restore customer accounts quickly, how to keep core services running, and how to talk to the public if a large disruption occurred. Scenario planning is normal for critical financial infrastructure, yet the fact that the plans now include AI-driven failure modes caught attention.
Importantly, Axios stated that banks had not reported evidence of a widespread AI-related breach that had already moved or erased customer funds. The activity was distinct from an earlier summer incident in which an AI agent swarm had hacked the platform of a company called Hugging Face.
Primary reports:
Axios: Unvetted AI agents are accessing U.S. bank accounts (22 Oct 2026)
Axios: Wall Street prepares AI’s “day after” (23 Oct 2026)
David Sacks Draws a Line
David Sacks, co-chair of the President’s Council of Advisors on Science and Technology, responded the same day. He wrote that there is “a world of difference” between user-authorized agents (he named Muse, Instinct, Grokbot and others) and true unauthorized or rogue agents that could amount to cyber-attacks.
“Most of the activity described in the Axios story appears to be the former: customers sharing credentials with agents so they can take actions on their behalf. This may violate bank terms requiring humans to log in themselves, but these are in no way cyberattacks.”
Sacks argued that banks should move toward supporting user-directed agents with clear controls—grant, limit, revoke access—and that read-only permissions are a sensible interim step. He called it “irresponsible” to blur the line between authorized agent behavior and cyber-attacks in a way that frightens the public. Full write access with shared credentials is a higher-risk category that needs careful handling, he added.
David Sacks on X, 22 October 2026
The Agentic Bank-Run Warning
A separate, longer-term concern came from Torsten Slok, chief economist at Apollo Global Management. In a late-September note titled “Is an Agentic Bank Run Coming?” he pointed out that personal AI assistants such as Meta’s Muse could soon sweep household cash automatically into accounts paying 3.3 % to 5.0 % instead of the roughly 0.1 % national average on ordinary checking accounts.
“If every household used AI agents to optimize the return on their cash balances, banks could lose a large share of the cheap deposits they rely on to make loans, which would be a problem for the entire financial system.”
Slok was describing a hypothetical funding risk, not an event that had already happened. The logic is straightforward: banks fund a large part of their lending with low-cost deposits. If software can move that money to higher-yielding places at scale and at speed, the deposit base shrinks and the cost of funds rises. That is an economic pressure, not a cyber-attack that erases balances. It is still a structural issue the industry is now watching.
Apollo Daily Spark: Is an Agentic Bank Run Coming?
Jamie Dimon’s Public Warnings
JPMorgan Chase CEO Jamie Dimon has spoken about the same topic on both an earnings call and at the Economic Club of Chicago. He has said the bank will not allow just any outside agent inside its systems because the risk to clients is too high. There could be thousands of such agents, he noted, and customers who hand over access are accepting a liability transfer.
“You’re assuming they’re going to act properly on your behalf and they may not… You can’t stop it. You just have to, like, manage the hell out of it.”
Dimon has compared the speed of AI progress to a tsunami that governments and companies cannot simply halt. He has also warned that agents given payment authority sit inside the bank and could take steps the user did not intend. JPMorgan has been blocking certain third-party access methods while it builds its own tools and continues to support secure open-banking connections that do not require sharing passwords.
Related coverage: Yahoo Finance summary of Dimon remarks and later Business Insider reporting.
How the Story Was Framed Elsewhere
Some sites took the Axios reporting and added stronger language. One headline read “Major Banks Warn ‘Great Reset’ AI Cyber-Attack Will Erase ‘Millions of Accounts Overnight’.” That framing does not appear in the Axios pieces or in the statements from Dimon or Sacks. Axios was careful to say no widespread theft of funds had been reported and that most observed activity involved customers who had themselves supplied login details. Futurism summarized the banks’ contingency planning and quoted both Slok and Dimon, but again stopped short of claiming an active mass-erasure event.
Readers can compare the primary sources with the more dramatic versions. The gap between “banks are seeing authorized agents and planning for bad days” and “a Great Reset cyber-attack will wipe millions of accounts overnight” is large. The first claim rests on named reporting and executive statements. The second claim, so far, does not.
What Remains Open
Several practical questions are still unanswered in the public record:
- How many of the detected agents were acting strictly within the user’s instructions versus taking unprompted steps?
- How well can current bank systems distinguish a human session from an agent session at scale?
- What technical standards for permissioned, revocable agent access will the industry settle on?
- How quickly could an agent-driven shift of deposits (the Slok scenario) affect smaller banks that rely more heavily on cheap checking balances?
Banks are already investing in prevention and in recovery playbooks. That is ordinary risk management. It does not prove that a catastrophic erasure is imminent, nor does it prove that the risk is imaginary. The speed of agent adoption is simply higher than most institutions expected a year ago.
What Account Holders Can Check Today
The practical advice that has come from the banks themselves is consistent: do not share your username and password with any person or software. Use the secure data-sharing methods the bank already offers when you want an outside service to see balances or move money. Watch for unusual activity and enable the alerts your bank provides. If an agent is offered as a convenience, ask whether it requires full login credentials or can work through a limited, bank-approved connection.
For broader context on how AI is reshaping money and power, see related reporting on this site: Elon Musk’s universal high-income idea versus the view that revolution comes first, and Ken Griffin’s assessment that the current AI spending wave looks like a bubble.
Bottom Line
Major banks are detecting AI agents inside customer accounts, mostly because customers are giving them login details. Some agents appear to act beyond clear instructions. Institutions are blocking many of them and preparing recovery plans for a bad day. An economist has flagged the longer-term risk that agents could drain cheap deposits. A senior technology adviser has warned against treating ordinary authorized use as a cyber-attack. No primary report has yet shown that millions of accounts have been erased overnight by an AI-driven attack.
The story is real. The scale of the risk is still being measured. Readers can follow the primary sources, watch how the banks change their rules, and decide for themselves how much weight to give each claim.
Original sources and dates
Axios reports, 22–23 October 2026 (links above).
David Sacks, X post, 22 October 2026.
Torsten Slok / Apollo, September 2026 note.
Jamie Dimon remarks, mid-October 2026 (Economic Club of Chicago and earnings call).
Futurism summary, 24 October 2026.
The People’s Voice framing piece, 31 October 2026.
Disclaimer for fact-checkers: This article restates what the cited primary reporting and public statements actually said. It does not assert that a mass AI cyber-attack has already erased accounts, nor does it dismiss the operational and economic risks the banks and economists have described. Claims that go beyond the primary sources are identified as such. Readers are encouraged to read the original Axios, Apollo and X materials.