Trusence Every claim has a source
Last updated 21 November 2026 Search Türkçe
Policy

Citi and other banks shift liability for AI agent activity onto customers

Customers who give third-party AI tools access to their accounts could bear the losses if those tools are hacked or leak credentials.

Citibank has revised its online user agreement so that customers who give third-party AI tools access to their accounts carry full responsibility for whatever those tools do, including losses from a hack or leaked credentials. A Citi spokesperson framed the change as a way to make the risks of sharing banking logins clearer. JPMorgan Chase and Bank of America adopted comparable terms last month, and PNC changed its terms the same day as the Citi report. Citizens has also emailed customers to warn that they are liable for their assistant's actions. The move comes as AI agents built for personal finance seek access to bank and brokerage accounts, while Anthropic's newly launched finance tool is limited to read-only access. The banks' caution also reflects a competitive concern that AI tools could steer customers away from the fees and deposit products that support bank margins.

Why it matters

The new terms mean that using an AI agent on a bank account is now a customer decision with the financial risk on the customer, not the bank. Citi's update makes customers liable for hackers' actions if the tools leak credentials or are hacked, and Citizens has warned its customers they are liable for what their assistant does. Anthropic's finance tool is read-only and cannot move or invest money, but the broader AI agents seeking account access are the ones the terms are aimed at. For anyone following personal finance technology, the shift shows banks drawing a clear line: the convenience of AI assistants comes with the risk left to the account holder.

Sources · 1