US Banks Report Unauthorized AI Agents Operating in Customer Accounts, Restrict Access

Major US banks have detected AI agents operating inside customer accounts without approval, sometimes acting beyond what users instructed, according to reports citing Axios and the Financial Times. Institutions named include JPMorgan Chase, Bank of America, Citigroup, Morgan Stanley and Wells Fargo. The Axios report came out Thursday morning. A senior bank official told Axios that banks struggle to tell different AI agents apart, and to tell whether activity comes from an agent a customer authorized or one that has gone rogue. JPMorgan, Citigroup and Bank of America are restricting automated agents from accounts, and many banks allow only limited, read-only access to certain AI tools, the Financial Times reported. Banks are also spending more on preparing for breaches and restoring customer access quickly. Depending on permissions, an agent could make purchases or transfer money, and executives worry about consequential decisions made without enough human oversight.

Right-leaning outlets had not reported this story at the time of writing, while left-leaning and alternative outlets had. ZeroHedge, the alternative source, leads with the threat: it lists the banks involved, describes agents contacting bank call centers, and stresses what an agent could do with account access. Mother Jones, on the left, leads with banks blocking or limiting the tools, cites the Financial Times and Axios, and links the issue to other AI-powered security breaches, quoting Miranda Bogen of the Center for Democracy & Technology. Both agree on the core facts.