Banks warn: AI agent mishaps are your problem

Original: More banks say customers using AI agents are on their own — even if they get hacked

Why This Matters

Five major U.S. banks setting liability precedent before federal AI rules exist.

Citibank, JPMorgan, Bank of America, PNC, and Citizens have all updated terms of service to disclaim liability for transactions—or hacks—caused by third-party AI agents customers authorize to access their accounts.

Citibank updated its Online User Agreement this week, joining JPMorgan, Bank of America, PNC, and Citizens in a coordinated industry shift: if a third-party AI agent touches your bank account and something goes wrong, you're on the hook. That includes unauthorized transactions, bad financial decisions, and even cases where the AI tool itself is hacked and credentials are stolen. 'We recently updated our Online User Agreement 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,' a Citi spokesperson told TechCrunch. Citi's terms go further than most—explicitly covering credential leaks from compromised AI tools. The five banks collectively represent the largest retail banking institutions in the U.S. by assets. The timing tracks the rapid growth of personal finance AI agents, which can analyze spending, offer investment advice, and execute transactions using stored credentials. Anthropic's recently launched personal finance tool took a notably cautious approach, billing itself as 'read-only' and explicitly stating it 'can't move, manage, or invest your money.' Banks have an additional motive beyond risk management: AI agents that help consumers dodge fees or find higher-yield accounts threaten the low-interest deposit float that underpins bank margins. Notably, Citi's new terms carve out an exception—AI agents provided directly by Citi won't trigger additional liability.

Source

techcrunch.com — Read original →