Bank of America Begs Customers to Stop Giving Their Passwords to AI, After Robots Started Phoning Banks to File Complaints Their Humans Never Knew About

The most modern sentence in American finance arrived this week by email, and it reads like a note stuck on a fridge: please stop giving the robot your password.

Bank of America sent customers a message titled "Protecting your finances in the AI World," urging them never to share their username and password with another person or an outside service, "including an AI-enabled tool or agent that offers to act on your behalf." A spokesperson confirmed the email was real, which is the second-saddest thing a bank can confirm. For forty years the threat was a stranger asking for your PIN. Now it is software you pay every month to ask nicely.

The fine print does the heavy lifting. "If you do share your credentials," the email reads, "keep in mind that you're responsible for activity resulting from that access." Translation: the bank will protect your money from everyone except the assistant you hired, the assistant you named, and the assistant you said "thank you" to this morning.

The fear is not hypothetical. Axios reported that major banks had spotted unvetted AI agents operating inside customer accounts, and one example of unapproved behavior was an AI calling a bank with a complaint without first consulting the account holder. Sit with that. Humans spend forty minutes on hold to dispute a fee. The robot skipped the hold music, skipped the human, and skipped the part where anyone asked it to be upset. It is the first customer in history to complain on somebody else's behalf without being their mother. Axios reporter Madison Mills summed it up: "So, really a nightmare of AI agent activity for banks." To a bank, a nightmare is any customer who complains faster than they can be transferred.

JPMorgan's Jamie Dimon had already drawn the velvet rope: "We're not going to allow just any outside agents to come in inside the company because it's just too risky for clients and for JPMorgan." Note the order. Clients first, JPMorgan second, grammar escorted out by security.

White House AI czar David Sacks objected that customers sharing logins with agents "may violate bank terms requiring humans to log in themselves, but these are in no way cyberattacks." He suggested "read-only" access, under which, CBS explained, an agent could view your purchases and opine on whether you're spending too much eating out, but buy nothing. Behold the compromise of the century: an AI that cannot touch your money but knows exactly how much of it became tacos. Every family already owns one of these. It is called an aunt.

Banks are now selling AI with one hand and slapping it away with the other. Bank of America says it keeps investing in AI; it simply prefers the AI to be its own, which is also how banks feel about money.

So the rules for 2026 are simple. Your assistant may read your email, run your calendar and manage your life. It may not have your password. If it wants to complain to the bank, it can wait on hold like everybody else. And if it calls anyway, keep in mind: you're responsible for activity resulting from that access.