Citi Shifts AI Agent Liability to Customers as Banks Push Back on Third-Party Finance Tools
Citibank, the third-largest commercial bank in the U.S., has updated its terms of service to make customers responsible for any activity carried out by third-party AI agents they allow into their accounts, including losses if those tools leak credentials or are hacked.
AI agents, from frontier lab models to dedicated personal finance apps, can be handed banking credentials to analyse spending, offer advice and even make transactions and bill payments. A Citi spokesperson said the bank wanted customers to understand the risks of sharing credentials with such tools, and that its updated Online User Agreement clarifies they are accountable for what those agents do. Citi-provided AI agents, however, carry no added liability.
The move follows similar policies from JPMorgan and Bank of America last month, with PNC and Citizens issuing comparable warnings this week.
Banks view AI agents as both a security risk and a competitive threat, since such tools could help consumers avoid fees and low-interest accounts. Some AI developers are treading carefully: Anthropic’s new personal finance tool is read-only and cannot move money. Similar resistance has emerged in ride-hailing and food delivery.