Bank of America Warns Customers: Don't Give Your Password to an AI

Bank of America has told customers in a mass email never to share their username and password with an outside service, "including an AI-enabled tool or agent that offers to act on your behalf." The message, titled "Protecting your finances in the AI World," marks one of the clearest signs yet that US consumers are handing banking credentials to AI assistants at a scale banks cannot ignore.

The warning carries a hard financial edge. Bank of America's email states the liability rule plainly: "If you do share your credentials, keep in mind that you're responsible for activity resulting from that access." Account agreements at most banks hold customers responsible for anyone they give their login to. If an AI agent drains funds or moves money badly, the loss sits with the customer, not the bank.

What can actually go wrong?

The risks the bank flagged fall into three buckets:

• An AI assistant could take action on your finances without your knowledge.
• The bank's fraud systems could interpret the agent's activity as an account takeover and lock you out.
• A security flaw in the assistant could leak your credentials, giving financially motivated hackers a direct route into your account.

The email itself reads as if written under strain. The tone suggests bank staff have already fielded cases where customers did exactly this, as personal AI assistants have grown more popular and users increasingly delegate logins to them.

Why are banks worried about AI agents now?

The warning lands while banks are still working out where AI agents fit in finance. According to reporting on the sector, banks are grappling with unauthorized agents accessing customers' money. The industry has no settled answer yet.

The topic surfaced on stage at TechCrunch Disrupt, where banks, payment providers, and processors discussed how to let customers reach their accounts through AI tools. One proposed model is agentic payments: customers set agents to autonomously buy things at the best price or at the right time. That model requires the industry to solve access, authorization, and fraud detection before it can scale safely.

Bank of America did not immediately respond to a request for comment about the email. The story was spotted and circulated by Bryan Toporek.

What does this mean for customers?

The mechanism here is simple. A password shared with an agent is a password shared with a third party under almost every bank's terms of service. The bank sees activity from an unknown actor. It cannot tell a helpful shopping agent from a thief. The customer absorbs whatever follows, whether that is a lockout or an unauthorized transfer.

The stakes rise as agent tools spread. Banks want a future where agents pay bills and hunt bargains on command. But until the industry builds dedicated rails for agentic payments, customers who improvise with shared logins do so without a safety net. Bank of America's message sets the boundary for now: the convenience of an AI assistant does not transfer the liability, and the bank has put that warning in writing.