SILICON VALLEY — Major U.S. banks are scrambling to head off damage from artificial intelligence agents, the AI assistants that carry out tasks on a user's behalf, after detecting signs that unidentified agents have been accessing their systems as customers share sensitive personal information with the tools. The arrival of Meta's Muse, OpenAI's Dots and Anthropic's Claude Duet has pushed agent use beyond the corporate clients who once dominated the market and into the general public, leaving commercial banks struggling to respond.
Bank of America, one of the largest U.S. lenders, sent customers an advisory email on the 26th titled "Protecting your finances in the AI World."
The email, sent under the name of Ashley Ross, Bank of America's chief client care executive, said the bank would adopt a range of AI tools but that security came first, and urged customers not to share personal information. "Bank of America continues to invest in innovative technology, including AI, to make our customers' financial lives more convenient, personalized and secure," the bank said, adding that it plans to introduce a wider range of money management tools but that customers should treat security as their top priority.
"Remember that you should never share your user name and password with anyone else, or provide them to outside services, including AI-based tools or agents that offer to perform tasks on your behalf," the bank said. It also said customers should keep in mind that if they share personal information, they are responsible for all activity resulting from that access. "We have high hopes for the possibilities that AI used responsibly will bring," the bank said. "We will do our best to protect the safety of our customers and their valuable assets while delivering innovative services."
Behind the largest U.S. bank's ban on sharing personal information — delivered with a warning that customers themselves bear responsibility for any losses — is the infiltration of bank systems by AI agents. U.S. online outlet Axios reported that banks had detected unauthorized third-party AI agents operating inside customer accounts, and had found that some of the agents were attempting tasks unrelated to user instructions.
Large banks including Bank of America, Morgan Stanley, JPMorgan and Wells Fargo say their rules and guidelines require users to access accounts themselves, and that delegating that access to agents creates considerable security risk. They are mounting an all-out effort to shut off any such exposure, with particular concern about incidents that could occur once users grant agents the authority to make purchases or transfer funds.
Responsibility for AI-related incidents remains poorly defined and penalty rules are not yet properly in place, making the risk of such incidents the biggest worry in the AI services market. Anthropic, the world's largest AI developer, went so far as to warn in the IPO registration statement it filed with the U.S. Securities and Exchange Commission the same day that liability and legal risks could become enormous and that insurers willing to cover them may be hard to find.
Banks have been tracking cases in which AI contacts call centers on a user's behalf or acts without user instruction, and see the possibility that agents could access customer accounts and affect their assets. Because banks face limits in distinguishing on their own between account holders and agents logging in, Morgan Stanley, JPMorgan, Bank of America and Citibank are restricting third-party agent access and rushing to revise their user agreements.
Banks also recognize that as personal AI agents become commonplace, more customers want to link them to their accounts. Lenders themselves are adopting AI features such as chatbots, mindful of cost savings and shifts in the market. The problem is that an incident caused by agent access raises the question of who is liable. For that reason, banks are blocking the sharing of personal information as much as possible while restricting personal agents such as Muse from accessing customer accounts and transferring funds. A Chase spokesperson urged customers not to give their account IDs and passwords to anyone.
