Agentic AI
Banks Shift Liability to Customers Using Third-Party AI for Banking
Major lenders including Citi, JPMorgan, and Bank of America are updating policies to make customers responsible for losses or breaches tied to AI agents managing their accounts.
TechCrunch — AI
Financial institutions are drawing a hard line on consumer use of third-party AI tools to manage banking. Citibank this week revised its terms of service to clarify that customers who grant AI agents access to their bank accounts assume all risk — including losses from hacking or accidental transactions executed by the automated systems.
The shift reflects growing tension between AI startups pursuing fintech opportunities and established banks wary of losing control over customer relationships and protecting themselves from liability.
The Risk Transfer#
Major U.S. lenders are moving in lockstep on the issue. JPMorgan and Bank of America announced comparable policies last month, and PNC—the eighth-largest bank—followed suit this week. Citizens Bank similarly notified customers via email of matching terms.
Citi's update goes further than simply disclaiming responsibility for AI-driven actions. The bank now explicitly states that if a third-party AI tool itself gets breached or leaks customer credentials, the customer bears responsibility for any unauthorized access or fraud that follows. In contrast, Citi's own AI tools carry no such additional liability for account holders.
Why Banks Are Pushing Back#
AI agents designed for personal finance can analyze spending patterns, suggest investment moves, and automate transactions and bill payments when granted access to banking credentials. While potentially useful, the arrangement makes banks uncomfortable with unpredictable AI behavior in their systems.
According to Citi, the update aims to "provide greater clarity about those risks and to reinforce that customers who choose to grant account access to third-party AI agents are responsible for activities conducted through those agents."
The resistance also reflects competitive concerns. If AI tools help consumers spot and avoid unnecessary fees or steer them toward better-rate accounts, that erodes traditional bank profit margins. By warning customers away and imposing liability, banks aim to protect their franchise while positioning their own AI offerings as the safer alternative.
A Cautious Approach Emerges#
Not all AI companies are rushing into unrestricted account access. Anthropic's recently launched personal finance tool exemplifies a more conservative model—it explicitly operates on a "read-only" basis, explicitly unable to move funds, make investments, or execute transactions.

