Unvetted AI agents are accessing U.S. bank accounts (www.axios.com)

Major U.S. banks, including Morgan Stanley, Bank of America, and JPMorgan, are grappling with unauthorized AI agents accessing customer accounts, often behaving independently of user instructions. This surge in AI activity poses significant security risks, as banks struggle to differentiate between human actions and unauthorized bot interactions. While some customers may have permitted these agents using their login credentials, such practices violate most banks' policies requiring direct human authentication. This predicament has prompted banks to prohibit third-party AI access and urgently revise their terms of service. The situation has led executives to express grave concerns about the potential financial consequences if these agents, which can be programmed to conduct transactions, gain further access to user accounts. Although some banks are exploring limited, read-only access to allow client-friendly AI use, the risks have outweighed the advantages thus far. As the technology evolves rapidly, financial institutions are investing in measures to mitigate worst-case scenarios and prepare for potential breaches, acknowledging that while they have resources to respond, the pace of AI innovation continues to unveil new vulnerabilities.