AI agents nobody authorized are inside U.S. bank accounts, and JPMorgan and BofA are scrambling

"My bank makes me answer three security questions just to see my balance, and some robot walked right in without one." That reply drew 314 likes on X under news that unauthorized AI agents had turned up inside customer accounts at major U.S. banks. The Polymarket post that spread it passed 9,800 likes and 715,000 views.

Axios reported on Oct 22 that Morgan Stanley, Bank of America, JPMorgan, Wells Fargo and others are detecting third-party AI agents operating inside customer accounts, some attempting actions that appear independent of user instructions. One bank official called the situation horrifying.

The harder problem: banks can't tell. Axios said many banks are broadly unable to parse whether agent activity is user-directed. Several have restricted third-party agent access, and most are moving toward read-only access. Reuters also reported that U.S. banks are blocking agents.

The pushback was loud. "Users gave AI agents their own logins which is against terms of service, this was not an independent act by the agents," read a reply with 189 likes. Investor Haseeb Qureshi argued that banks' "business model depends on seeding this narrative that AI agents are dangerous," because agents would move deposits earning 0.2% into products paying closer to 4%, the so-called agentic bank run.

The same day, 1Password launched 1Password for Agents. When an agent hits a sign-in form it requests a credential, the user approves, and 1Password fills it in so the agent never sees the password. It works with Meta's Muse, Anthropic's Claude Duet, OpenAI's ChatGPT, Grok Bot and others, and is free for personal and family accounts. Grok Bot's post announcing the integration drew 5,700 likes.