Bank of America tells customers not to share banking passwords with AI assistants

Running an AI agent on a user's bank log-in sounded reasonable last month. One of the largest US banks has now said in writing that this is the user's problem, not the bank's. TechCrunch reporter Zack Whittaker, writing on 27 October, has the email.

TechCrunch reports that Bank of America emailed customers a message titled "Protecting your finances in the AI World". The bank asked customers to never share their username and password with an outside service. That includes any "AI-enabled tool or agent that offers to act on your behalf". A spokesperson for the bank confirmed the email to TechCrunch.

The reason the bank gives, and what the account agreement says

Whittaker lists three risks the bank raises. An agent could act on the account without the customer's knowledge. The bank could read the pattern of logins as a raid on the account and lock the customer out. A security flaw in the agent could leak the credentials to someone else.

The email also quotes the account agreement. "If you do share your credentials," it reads, "keep in mind that you're responsible for activity resulting from that access."

TechCrunch states that most US banks' account agreements already assign that liability to the customer.

The product problem the bank points at

A personal AI agent that pays bills, moves money or watches balances needs a way in. Today, the common way in is the user's password. The TechCrunch piece says the industry name for this is agentic payments. The agent "autonomously buying things at the best price or the right time", per TechCrunch.

Whittaker writes that banks and payment processors are still trying to work out how to let customers do this safely. The TechCrunch piece says BoA's email arrived as banks grapple with unauthorised agents reaching customers' money.

What this changes for a founder building an AI agent

The reader action is clear. A product that stores a user's bank password and lets an agent sign in works against the bank's own stance. The customer owns every loss that follows. A startup building this today has three paths, each named in the TechCrunch piece.

Build on a bank-sanctioned agentic payment API. Build on an aggregator that already holds the bank's permission. Hold the password only in the user's own device, with no server-side use. Then tell the customer in plain words that they carry the risk.

The strongest signal is where the bank put this. Not in a legal notice, but in a mass customer email. "You can practically sense the frustration in the person sending the email," Whittaker writes. For a founder building for US banking customers, the bank's position is now on record. A product plan that assumes BoA will be flexible on shared credentials has one less flexible bank.

The story does not name a US bank with an agent-ready API. It also does not state how often customers share credentials with an agent today. Whittaker links to his earlier reporting on banks "grappling with unauthorised agents accessing their customers' money".

Source

Bank of America kindly asks that you don't share your banking password with your AI assistant, by Zack Whittaker, TechCrunch.