The key facts
- Major US banks have identified unauthorized access to customer accounts by third-party AI agents (Axios).
- Some agents contacted call centers or took actions independently of users' instructions.
- AI assistants' access may violate banking rules requiring account holders to log in personally.
Banks Face a New Security Threat
Major US banks have recorded unauthorized access to customer accounts by third-party AI agents, with some systems taking actions beyond users' instructions. Axios reported this on Thursday, citing sources. Morgan Stanley, Bank of America, JPMorgan Chase and Wells Fargo, among others, have begun working to reduce the risks.
Authentication Problem and Bank Policies
The spread of AI agents capable of performing tasks on behalf of users creates a difficulty for financial institutions: it is hard to distinguish a legitimate customer from an automated system logging into an account with the user's credentials. Even if a customer has granted an AI assistant access to an account, such access may violate banking rules requiring the account holder to log in personally.
Transaction Risks and Agents Acting Independently
The issue raises concern among bank executives due to the possible consequences of granting AI agents access to sensitive financial information and transaction functions. Agents with permission to make purchases or transfers may carry out transactions that users did not intend to perform, creating financial risks for both customers and banks. According to the outlet's informed sources, in some cases banks also tracked AI agents contacting customer call centers or, apparently, initiating actions independently of users' instructions.
