Axios: US banks restrict AI agents accessing customer accounts
🎧 Listen to this story
According to Axios, cited by Xataka, some customers in the US are allowing AI agents to access their bank accounts using their credentials. Banks fear that a misinterpreted instruction could lead to purchases or money transfers, and say they have observed agents contacting customer-service centers or taking actions that appear independent of the user’s instructions—although it is not yet clear whether they actually acted on their own.
Axios reports that Morgan Stanley has begun blocking third-party agents, while JPMorgan, Bank of America and Citi are also imposing restrictions. Wells Fargo will include AI agents in terms taking effect on November 24, covering unwanted actions, misinterpreted instructions and situations in which an agent exceeds the authority granted by the customer. Banks are also preparing to quickly restore account access if something goes wrong.
Why it matters · editorial interpretation
The restrictions reported by Axios could make it harder to use agents for banking tasks and pressure customers, banks, and providers to define clear authorization limits and rapid recovery procedures when something goes wrong. Uncertainty about allegedly independent actions reinforces the need to distinguish user instructions from an agent’s interpretations.
Sources
SecurityAgents