Major U.S. banks are detecting unauthorized activity from third-party AI agents operating within their customers' online banking accounts, according to reporting by Axios. Some of the largest financial institutions in the country say customers have provided their personal AI agents, which the banks have not vetted, with access to their financial accounts. In many cases, customers are sharing their usernames or passwords with AI agents in violation of their banks' terms of service. In other instances, banks found evidence of agents taking actions in customer accounts that users had not explicitly authorized.
Examples of unauthorized agent behavior include AI calling a bank on behalf of a customer to lodge a complaint without first consulting the account holder. However, these agents were not hacking into accounts or depleting customer funds. Instead, they were largely following user instructions without the bank's permission to transact on behalf of account holders. The issue is industrywide, affecting dozens of financial institutions as they move to restrict AI agents' access to customer accounts.
JPMorgan Chase CEO Jamie Dimon raised concerns about AI agents taking autonomous actions in customer accounts during the company's earnings conference call on October 13. Chase said in a statement that customers should not share their usernames and passwords with anyone, and that the bank supports secure, customer-authorized data sharing through open banking providers. Charles Schwab said it has been blocking unauthorized agents, according to Reuters. Morgan Stanley said it had not observed any unauthorized AI agent activity across its platforms. Bank of America and Wells Fargo did not respond to requests for comment.
White House AI czar David Sacks responded to the report by distinguishing between user-authorized personal AI agents that help manage bank accounts and unauthorized agents used for cyberattacks. He said most of the activity described in the Axios story involved customers sharing credentials with agents to take actions on their behalf, which may violate bank terms but does not constitute cyberattacks. Sacks suggested that banks could grant AI agents "read-only" permission, allowing them to view a user's spending and account activity but preventing them from taking autonomous actions. He noted that giving full write access with user credentials represents a higher risk category that should be approached carefully.
For US savers and borrowers, this means banks are working to restrict how personal AI agents can access and operate within their accounts to reduce security risks, even as customers increasingly look to automate their financial management.