Bank of America warns customers not to share banking passwords with AI agents

Bank of America has begun emailing customers with a pointed piece of security advice: do not give your online banking username and password to an AI assistant.

According to TechCrunch, which reported the notice on October 27, the message was sent under the heading "Protecting your finances in the AI World" and tells customers to keep their log-in details away from any other person or outside service, explicitly including AI-enabled tools and agents that offer to act on a user's behalf. The bank did not immediately respond when TechCrunch requested comment.

A warning aimed at a growing habit

The notice is not a response to a disclosed breach. As TechCrunch describes it, the bank is reacting to behaviour that is becoming more common as personal AI assistants spread: users handing over banking credentials so the software can check balances, move money, or otherwise manage their finances.

That arrangement creates several distinct problems, the report notes:

• The agent may take actions in the account that the customer never intended or approved.
• Automated log-ins can resemble an account takeover to the bank's fraud systems, which may respond by locking the customer out.
• A security flaw in the assistant itself could expose stored credentials to financially motivated attackers, handing them a direct route into the account.

The liability lands on the customer

The sharpest part of the warning concerns consequences. Account agreements at most banks, TechCrunch reports, make customers answerable for anything done by people or services they voluntarily give access to. Bank of America's email makes the point explicit, stating that anyone who shares credentials is "responsible for activity resulting from that access."

In practice, that means the protections a customer might normally expect after unauthorised transactions may not apply when access was handed over willingly, even to a tool the customer believed was acting in their interest.

Banks are still working out agent access

The warning also lands in the middle of a broader industry conversation. According to TechCrunch, banks, payment providers and processors are still figuring out how AI agents should be permitted to interact with customer accounts at all. One concept under discussion is agentic payments, in which an agent is authorised to buy things on a user's behalf, timing purchases or hunting for the best price autonomously, a topic explored onstage at TechCrunch's Disrupt event.

Until sanctioned mechanisms like that exist, raw credential sharing is the workaround, and banks are reportedly already grappling with unauthorised agents reaching into customers' money.

Why it matters

A mass email from one of the largest banks in the United States is a strong signal that sharing log-ins with AI agents has moved from hypothetical to common enough to warrant a warning. It exposes a mismatch between the pace of consumer AI adoption and banking security models designed around a single human password holder.

The stakes are asymmetric: if an agent misfires, a credential leaks, or a fraud lockout hits, the customer absorbs the consequences under standard account terms. That dynamic puts pressure on banks and AI providers alike to build controlled, auditable ways for agents to act on accounts, because in the absence of official channels, users will keep filling the gap with their passwords.