Major U.S. banks detected unauthorized AI agents inside customer accounts. Some agents acted without user instructions. Banks are blocking third-party agent access. Morgan Stanley, JPMorgan, Bank of America and Wells Fargo are restricting agents. Most banks now allow only read-only access for agents. Executives call the situation horrifying. Banks struggle to tell user-directed agents from rogue ones. Some agents contacted call centers independently. Banks are investing in worst-case scenario planning. The rapid pace of AI advancement is driving the security concerns.