Citibank updated its online user agreement to say customers who give third-party AI tools access to their accounts are responsible for activity conducted through them. Its terms also cover credential theft: if an AI tool leaks a customer's credentials or is hacked, the customer is responsible for what hackers do with the account, even if the customer did nothing wrong. Customers can use Citi's own agents without extra liability. JPMorgan and Bank of America announced similar policies in October, PNC changed its terms on 18 November, and Citizens emailed customers warning they are responsible for anything their AI assistant does.
The banks say they are wary of unreliable AI actors in their systems and will not reverse transactions made by automated assistants, intentionally or not. They also see AI tools as a competitive threat, and markets briefly fell after reports that AI tools could help consumers avoid fees or low-interest deposit accounts that support bank margins. AI agents can be given credentials to bank and brokerage accounts to analyse spending, advise and pay bills.
Anthropic launched a personal finance tool in early November that it describes as read-only and unable to move or invest money. Service providers have pushed back on AI agents in other sectors, including ride-hailing and food delivery. No regulators or specific hacking incidents were named.





