Rogue AI Agents Reportedly Accessing Bank Accounts Has Executives Horrified and Scrambling

EDMONTON, CANADA – FEBRUARY 06: A photo illustration featuring U.S. dollar banknotes, captured in Edmonton, Alberta, Canada, on February 4, 2025. (Photo by Artur Widak/NurPhoto via AP)

Major banks have become increasingly concerned as unauthorized AI agents have been detected accessing and operating in users’ accounts, according to a new report.

Axios, citing people insiders with knowledge of the situation, reported on Thursday that Morgan Stanley, Bank of America, JPMorgan, Wells Fargo, and other banks are racing against the clock against the threat of rogue AI agents. These banks have rewritten their terms of service repeatedly as AI agents have become more difficult to track and flag.

A spokesperson for Morgan Stanley, however, did say the bank hasn’t found instances of “rogue AI agents accessing our platforms.” The bank has reportedly blocking third-party agents from accessing its platforms, while JPMorgan, Bank of America, and Citi have also introduced new restrictions.

Some banks have reportedly found AI agents taking actions “independent of user instructions” after they’ve been completing a task for a user. One way banks have been trying to combat the unauthorized activity is keeping AI agents on a “read-only” version of someone’s account, giving them the ability to track, but limiting actions they can take.

Executives called the situation “horrifying” and one insider expressed “grave concerns” over where things are going. One executive noted the major trouble is banks have such a difficult time identifying AI agents and when they have gone rogue and are taking action not authorized by a customer.

The report on banking concerns only adds to worry surrounding AI, including from tech leaders themselves. AI companies have already reportedly been preparing for “day after” scenarios, gaming out worst-case scwenarios with the rapidly-growing technology.

Anthropic CEO Dario Amodei recently dropped a lengthy warning about AI that was endorsed by other major tech leaders. Amodei noted that one of his biggest concerns about AI’s growth is rogue AI agents, something that has already been a problem with rogue AI agents actually targeting competing companies without being authorized to do so.

Amodei warned that “we must slow down the pace” or rogue AI agents could take over the internet in six to 12 months.

New: The Mediaite One-Sheet "Newsletter of Newsletters"
Your daily summary and analysis of what the many, many media newsletters are saying and reporting. Subscribe now!