US Banks Block Unauthorized AI Agents Accessing Client Accounts

Major US banks are discovering unauthorized AI agents built by third-party developers that are accessing client accounts, according to Axios. Some of these agents are attempting actions that appear to fall outside what users instructed them to do.

The arrival of such agents has become a serious problem for lenders, the outlet reports, because they are being woven into the financial system faster than protective mechanisms can be developed and deployed.

Telling human activity apart from an unauthorized bot is pushing banks to block the technology and quickly revise the terms of their services, according to the report.

Several large US banks, including Morgan Stanley, have begun blocking third-party AI agents from their platforms. JPMorgan, Bank of America and Citi are also introducing restrictions on their use.

Broadly, banks want to allow such agents to reach client data only in read-only mode, which permits retrieving financial information but not carrying out transactions.