AI

Wells Fargo’s new terms make your AI agent’s mistakes your problem

Susan Hill
Add us on Google

Wells Fargo is rewriting the agreement behind its app and website so that anything an outside AI agent does in a customer’s account counts as the customer’s own act. The bank says it will follow an agent’s instructions without checking them, will not watch what the agent does, and will treat the agent’s mistakes as the customer’s. “Your use of any AI Agent is AT YOUR OWN RISK,” the new terms say, in capitals.

For anyone who has handed a banking password to a shopping bot or a budgeting assistant, that changes who pays when the software gets it wrong. A double payment, an order placed ten times or a transfer triggered by a booby-trapped web page would sit with the account holder, not the bank, unless a law says otherwise. Nobody has to sign anything: customers who keep using online banking after the change are taken to have accepted it. “Consumers authorize goals, not every click,” Kacy Kim, a marketing professor at Bryant University who studies consumer behavior, told Wired.

The definition is wide. Wells Fargo calls an “AI Agent” any software that takes “autonomous, semi-autonomous, or non-autonomous action” on someone’s behalf, which reaches well beyond the chatbots sold as personal assistants. The customer carries the cost of an agent misreading instructions, doing things nobody asked for, acting on “manipulated inputs” or going beyond the authority it was given. Agents run by people a customer lets into the account fall under the same rules, and nobody under 18 may connect one at all.

Switching an agent off is also the customer’s job. The terms say customers must revoke its access themselves, keep evidence that they did and hand that evidence over if the bank asks. Until they tell Wells Fargo, they stay responsible for whatever the agent does. That follows the logic of the US rule on electronic transfers, Regulation E, which treats transfers by someone you gave your card or login to as authorized until you tell the bank to stop them.

Wells Fargo still lets agents in, which puts it at the permissive end of an industry that is closing the door. JPMorgan Chase changed its terms to ban third-party agents that log in with a customer’s credentials, Bank of America warned customers not to share their logins with agents, and Citibank and PNC added warnings of their own, according to Wired. Morgan Stanley has blocked third-party agents from its platforms, Axios reported. Wells Fargo has kept the right to cut off the connections agents use or block any agent outright.

The pushback reaches beyond banking. Amazon has blocked Meta’s Muse agent from making purchases on its store, and United Airlines has said personal bots will not be able to buy flights for travelers. American Express is the rare exception in Wired’s account: it said some agent mistakes, such as ordering too many of an item, may qualify for reimbursement in the future if the shopper cannot get the money back from the retailer.

How much the new wording changes in practice is unclear. Its clauses repeatedly defer to “law or regulation,” and no court has ruled on who carries the loss when an agent a customer switched on does something the customer never intended. The audience is also small. “All of our data shows that most people don’t even know what agents are, much less use them,” Forrester analyst Sucharita Kodali told Wired. And the banks are not against agents as such. “They’re locking the door to outside agents and building their own,” said Luca Cian, a University of Virginia professor who studies consumer psychology and AI. Wells Fargo’s own generative AI assistant has handled more than a billion interactions.

The new terms only cover customers of Wells Fargo’s US online and mobile banking, and every bank named so far is American. In the European Union, payment rules already push apps that read or move money toward licensed connections supplied by the bank rather than a shared password, which leaves less room for agents that log in as the customer.

The updated Wells Fargo agreement takes effect on Tuesday, November 24, and applies to anyone who keeps using the bank’s website or app after that day. JPMorgan’s ban on credential-based agents arrived in late October, the same day Bank of America sent its warning. Wells Fargo has also kept the option of making customers register their agents with the bank, which would give it a list of the bots it says it does not watch.

Tags: , , , , ,

Add us on Google

Discussion

There are 0 comments.