Somewhere inside Bank of America, someone had to write an email telling grown adults not to give their banking password to a chatbot. You can practically feel the sigh. But the message is worth reading closely, because it says a lot about where liability lands when AI agents start touching your money.
In an email blast to customers titled “Protecting your finances in the AI World,” the bank urged people to never 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 Bank of America spokesperson confirmed the email is real after it circulated on social media.
The reason for the warning is simple. Personal AI assistants are getting popular, and plenty of users are handing over log-ins so the agent can check balances, move money, or pay bills. That is a bad trade. An agent could do something with your finances without your knowledge. Your bank could see odd automated activity, assume someone is raiding the account, and lock you out. And if the assistant has a security flaw, it could leak your credentials somewhere else, which gives financially motivated hackers a direct way in.
That last risk deserves more attention than it gets. Most AI assistants are built by companies whose data handling you cannot audit. Your password may sit in logs, memory features, or third-party plugins. You are trusting a vendor’s security promises, and as anyone who follows breach reports knows, those promises have a poor track record.
Here is the part that should bother you. Bank of America says that if you share your credentials, “you’re responsible for activity resulting from that access.” That mirrors account agreements at most banks, which hold customers liable for anyone they give a log-in to. So the usual fraud protections you count on may not apply if an agent, or whoever compromises it, drains your account. The bank is not just giving advice. It is putting you on notice.
There is also a bigger story here. Banks, payment providers, and processors are still figuring out how to let customers use AI tools with their accounts in a safe way. One example is agentic payments, where you tell an agent to buy something at the best price or at the right moment, and it does so on its own. Banks are reportedly struggling with unauthorized agents already accessing customer money. As discussed onstage at TechCrunch Disrupt, nobody has settled how to do this properly.
The proper answer is scoped, revocable access. Think tokens with limited permissions, spending caps, and clear audit trails, not a master password pasted into a prompt. Until banks offer that, customers who want automation will keep reaching for the dangerous shortcut, because it works. And the bank will keep sending emails like this one.
So what should you do? Treat your banking password like a house key. Don’t give it to an AI tool, no matter how helpful the pitch. If a service demands your credentials to function, that tells you it was not built with your security in mind. Walk away, and let the banks build the safer rails first.