Four US banks stop AI agents that use customer passwords

Between 13 and 20 October 2026, four large US banks moved against AI agents that use the password of a customer on an account. A law firm update tells that JPMorgan Chase, Charles Schwab and Citigroup stop them. Wells Fargo changed its online terms. The customer has the risk of agent use. None of the four stops all AI use.

Why banks act

Apollo economist Torsten Slok warns that agents can move cash from accounts that pay about 0.1 percent to accounts that pay 3.3 to 5 percent. One estimate puts the cost to banks at $47 billion each year.

An agent with a password looks the same as the customer to the bank. It has the same rights, no limits and no audit trail.

The proof

The update gives no count of fraud losses. It tells that fake agent phishing campaigns are in use. The $47 billion figure is an estimate.

What is not known

The US consumer finance regulator wrote an open banking rule in October 2024. A court stopped it. A new version went to government inspection in August 2026 and is not published. It will select the fees for data access and who is an approved agent.

Insight: Banks guard their deposits as much as their logins

The estimate of $47 billion each year is the cost of interest on cash that gets no interest at this time. A different audit found that 37 of 51 US retailers stop agents by accident. This reads as a move to keep the customer and to set the price of the access that agents must use. The open banking rule on fees can select who has more control. Builders must plan to pay for access.