Hot AI News

Wells Fargo’s new terms put AI agent mistakes on the customer

Wells Fargo's updated online agreement, effective November 24, treats an AI agent's transactions as authorized by the customer, while JPMorgan, Bank of America, Citi and PNC also tighten their rules.

Wells Fargo’s new terms put AI agent mistakes on the customer
Photo: Lumen Wilde / Wikimedia Commons, CC BY 2.0

Wells Fargo will put a revised online agreement into effect on November 24 that treats transactions made by a customer’s AI agent as the customer’s own, and leaves the customer to supervise the agent. Wired reported that other large US banks are moving the same way.

What the Wells Fargo agreement says about agents

The bank’s updates page says the revised Online Access Agreement takes effect on November 24, 2026, and that customers do not need to take any action. The version now in force is dated May 21, 2026. The upcoming text, which Wells Fargo has published, refers to an AI Agent as something an Authorized Representative or Delegate may use.

A Wells Fargo branch at night with a glowing yellow sign above two lit walk-up ATMs set into a brick wall.
Photo: Ildar Sagdejev ( Specious ) / Wikimedia Commons, CC BY-SA 4.0

Under that text, the transactions of those agents count as authorized by the customer unless the customer tells the bank they are no longer authorized and the bank has had a reasonable opportunity to act on that notice. The bank may also rely on instructions from a delegate’s AI agents. On supervision, the draft is blunt:

You are solely responsible for supervising all of your Authorized Representatives and Delegates (including their AI Agents)

Wired quotes the new terms as saying that use of any AI agent is “AT YOUR OWN RISK”. Wells Fargo’s updates page adds that a third-party AI tool is not operated by the bank, that requests made through it may be treated as coming directly from the customer, and that information shared with it falls under the tool provider’s privacy terms. The bank also reserves the right to restrict or block access from certain third-party AI tools.

The overview of the upcoming agreement lists customer responsibilities for AI agents “if and when we permit the use of such agents”, which signals that permission is the bank’s to give or withdraw.

Advertisements

Other banks are drawing the same line

According to Wired, JPMorgan Chase changed its terms in late October to ban third-party agents that log in with a customer’s credentials. Bank of America warned customers the same day not to share login details with agents. Wired adds that Citibank and PNC have also updated their terms with agent warnings. We have not independently read those banks’ documents, so those details rest on Wired’s account.

The shift is not limited to banking. Wired says Amazon has blocked Meta’s Muse from making purchases on its platform and that United has said personal bots will not be allowed to buy flights for customers. American Express, Wired reports, has said some agent errors, such as ordering too much of an item, may become eligible for reimbursement if the customer cannot recover the money from the retailer. That is an announcement of future intent, and the terms are not described.

Liability follows the credential, not the intent

The angle that matters for readers is who carries the loss. The Wells Fargo text does not ask whether the customer intended a purchase. It asks whether the customer’s delegate, human or software, acted. A customer who gives an agent access has, in the bank’s drafting, authorized what the agent does until the bank is told otherwise and has had a reasonable chance to act.

That is why the revocation step carries weight. The text we read names a phone line for managing delegate access and a way to remove online access for an account. We did not find AI-specific revocation steps in the portions we could read, and part of the upcoming agreement could not be reviewed in full, so customers should read the published agreement itself.

The model also explains the bank warnings about passwords. Handing a username and password to an outside tool is the opposite of scoped access. Our explainer on OAuth scopes describes the alternative, in which an app receives limited, revocable permission without the password. Whether banks will offer that route to agents is not stated in anything we read.

Wired quotes a Forrester analyst who says most people do not know what agents are, and describes banks as risk-averse by nature, so early rules may be precaution rather than a response to widespread losses. Wired also notes that banks are building their own agents even as they close the door to outside ones.

Advertisements

What is not established

Wired’s reporting does not say how many bank customers use agents or how many have lost money through them. The wording of the JPMorgan, Bank of America, Citi and PNC changes is also not quoted in full. For Wells Fargo customers, the practical step before November 24 is to check whether any outside tool holds their login, and to revoke it if so.

Sign Up for the Daily AI Pulse

One email a day. All the stories that matter.

Alex CarterAI News & Big Tech Correspondent

Alex Carter writes for Aiholics, keeping readers updated on the fast-paced world of AI and Big Tech.

Leave a Reply

Your email address will not be published. Required fields are marked *

You may also like