THE CRUNCH

Citibank has updated its terms of service to say that customers who give third-party AI tools access to their bank accounts are responsible for whatever those tools do, even if the tool is hacked and the credentials leak. The change, reported on Wednesday, means Citi will not cover reversing transactions made by automated assistants, whether intentional or not. A Citi spokesperson said the Online User Agreement update was meant to clarify the risks of sharing banking credentials with third-party AI agents. Citi is the third largest commercial bank in the United States, and it is not alone: JPMorgan and Bank of America announced similar policies last month, PNC changed its terms the same day, and Citizens emailed customers this week with a comparable warning, according to an email seen by TechCrunch.

AI agents, from frontier lab models to purpose-built personal finance tools, can be handed credentials to bank accounts and other financial portals to analyse spending, offer advice and even handle transactions and bill payments. Banks are notably unenthusiastic about these unreliable actors in their systems, and the new terms shift the consequences to the people who granted access.

There is a competitive edge to the pushback too. Markets briefly wobbled at the prospect of AI tools helping consumers dodge fees and low-interest deposit accounts that pad bank margins, and banks will want to control that pathway if agents become a common way to manage money. Citi's terms note that its own AI agents can be used without additional liability.

AI startups are not standing still. Anthropic's personal finance tool, launched earlier this month, is explicitly billed as read-only and cannot move, manage or invest your money, a telling acknowledgment of how hard the trust problem is when the alternative is being on the hook for everything an agent does.