Axios reported Oct 22 that major U.S. banks are detecting unauthorized third-party AI agents inside customer accounts, some attempting actions that look independent of user instructions, per people familiar with the matter, as summarized by Cedar News. Most "unauthorized" cases were customers sharing their own logins with an agent the bank never approved, which still violates bank rules, with banks unable to tell customer-sent agents from rogue ones, and Axios gave no case count and no losses, per @shanaka86 summary. Named as restricting access are Morgan Stanley, JPMorgan, Bank of America and Citi in Coin Bureau's post, while Cedar names Morgan Stanley, Bank of America, JPMorgan and Wells Fargo. Morgan Stanley is described as blocking outside agents but saying it has not seen 'rogue AI agents accessing our platforms', with Citi, JPMorgan and BofA restricting access. The reported risks are agents moving funds or buying without clear authorization, confusion between human and bot activity, and agents contacting call centers for customers, pushing banks toward read-only access. No bank X account had posted on it by Oct 22, and I found no bank filing tied to this story.