Major U.S. banks are moving to restrict third-party AI agents after detecting activity inside customer accounts that they cannot confidently classify as customer-directed or rogue.

The alarm grew as banks found agents logging in with customers’ own credentials — often with permission, but in conflict with rules requiring account holders themselves to sign in. Executives also reported cases in which agents contacted call centers or appeared to act beyond a user’s instructions, raising fears over transfers, purchases and other financial decisions. One official described the industry’s concerns as “grave.”¹

Morgan Stanley has begun blocking third-party agents from its platforms, according to Axios, while JPMorgan, Bank of America and Citi are also restricting them. Chase has told customers not to share usernames and passwords; Morgan Stanley said it had not observed “rogue AI agents accessing our platforms.” The practical problem is broader: banks say they struggle to distinguish one agent from another — and to determine whether an agent was authorized or gained access improperly.¹

For now, the emerging compromise is read-only access, similar to permissions already granted to budgeting apps. But banks say the risk of allowing tools to execute transactions is too high while the technology evolves faster than their controls.

David Sacks pushed back on the framing of the reports, arguing that the crucial divide is not between humans and all bots, but between authorized assistants and malicious intruders. “There’s a world of difference between user-authorized agents … and true unauthorized/rogue agents that could constitute cyberattacks,” he wrote, adding that most activity described by Axios appeared to fall into the former category.David Sacks (DavidSacks)²

That disagreement now sits at the center of the banking response: institutions see uncertainty itself as a security threat, while AI advocates warn that indiscriminate blocks could punish customers for using legitimate digital help.