
Citi has rewritten its online banking agreement in the United States so that customers who let an outside AI agent into their accounts carry the risk of what happens next. An AI agent is a program that acts for a person, for example by logging in to pay bills or move money. A new section added in November warns customers against using such agents and says Citi will treat anything an agent does with their login details as authorized by them. If those details are later leaked or passed on to other people or programs, Citi will treat that access as authorized too, and it says it has no duty to check whether the agent followed the customer’s instructions. Citi’s own AI tools are exempt.
TechCrunch reported on November 18 that PNC changed its terms that day and that Citizens emailed customers a similar warning, while Bank of America told customers in October that sharing a login with an AI tool makes them responsible for the results. According to Decrypt, PNC’s new rules forbid giving an agent passwords or one-time security codes. Decrypt also notes that federal consumer rules require banks to investigate unauthorized electronic transfers, and that these protections cannot simply be signed away in an agreement. How those rules apply to an agent the customer invited in is still unclear, so the banks’ own advice for now is plain: do not give an AI assistant your banking password.