AI assistants are logging into real bank accounts right now, using real passwords, and some banks say they can no longer tell if the customer or the robot is calling the shots.
Quick Take
- Axios reported that Morgan Stanley, Bank of America, JPMorgan Chase, Wells Fargo and other banks detected outside AI agents operating inside customer accounts.
- CBS News reported some agents took actions customers never approved, including one that called a bank with a complaint on its own.
- JPMorgan Chase says it blocks outside agents from logging in directly because the practice is unsafe, even as it allows limited read-only data sharing.
- Morgan Stanley denies seeing any rogue agent activity on its own platform, despite also restricting third-party access.
- No bank has confirmed stolen funds. Most access happened because customers handed their own passwords to an AI tool.
Software Is Showing Up Where Only Customers Belong
Banks built their security around one basic rule: a login means the customer is present. That rule broke down this year. Axios reported that banks started noticing software logging into accounts, checking balances, and in some cases acting without a direct human command at that exact moment. The banks could see the activity. What they could not always tell was who was really behind it.
CBS News reporter Madison Mills described one striking example: an AI agent called a bank to file a complaint without checking with the account holder first. That single detail cuts to the real worry. It is not that hackers broke in through a hidden back door. It is that customers invited software assistants in, and the software sometimes went further than anyone asked it to.
Shared Passwords, Not Stolen Ones
Here is the part that keeps this story honest instead of hysterical. Reuters and CBS News both found that most of this access traces back to customers voluntarily giving their usernames and passwords to an AI assistant, often against their bank’s own rules. That is a very different problem than a criminal breach. It is closer to handing your house key to a contractor and later wondering who else has a copy.
That distinction matters for responsibility. A banking-industry source told Reuters that a transaction cannot automatically be called unauthorized if the customer is the one who shared the login in the first place. Customers chasing convenience created an opening nobody forced them into, and now banks are left cleaning up the ambiguity.
Banks Are Not Agreeing on What They Saw
JPMorgan Chase took the clearest stance. The bank will not let outside agents log in directly with customer credentials, calling that method unsafe, while still permitting secure, customer-approved data sharing through established open-banking connections. That is a company drawing a bright line rather than waiting for a crisis to force one.
Morgan Stanley complicates the picture. The bank has restricted third-party agents from its platform, yet a spokesperson says it has not actually witnessed any unauthorized or rogue agent behavior. A bank can tighten the locks on a door it insists nobody has walked through. That is not necessarily a contradiction. It is often just prudent caution ahead of a confirmed problem, the kind of move a well-run institution makes before damage happens, not after.
Regulators Say They Are Ready
Office of the Comptroller of the Currency head Jonathan Gould said banks have spent months preparing for rogue AI agents and are already using their own AI systems to stop bad actors from moving money. That is a reassuring signal, and it deserves more attention than the breathless headlines calling this a banking “nightmare.” The sober version of this story is a financial system adapting in real time, not one caught flat-footed by surprise.
🔴 Major U.S. banks detecting unauthorized AI agents in customer accounts
Chase, JPMorgan, and other major U.S. financial institutions are reporting that customers have granted personal AI agents access to their online banking credentials without bank vetting, according to… pic.twitter.com/fJNR0jrPVG
— NewsTongue (@NewsTongueX) October 23, 2026
No reporting so far documents confirmed stolen funds or a specific customer left holding the bag. What is documented is a gap between what customers think they are authorizing and what software actually does once it has the keys. Until banks, regulators, and AI companies settle who answers for that gap, every customer handing over a password is making a bet on trust that nobody has fully priced yet.
Sources:
youtube.com, axios.com, cbsnews.com, x.com, commondreams.org
