Our AI Agents A Freaking Out Banks. I’m sorry, the agents are relentless in junk fees. Major U.S. banks are detecting unauthorized third-party AI agents operating inside customer accounts, sources familiar with the issue told Axios on October 22. Some of those agents appears, according to the bankers, attempted actions that appear independent of user instructions. The access usually begins when a customer shares login credentials with an AI assistant or agent. The agent then logs in as the user. Most bank guidelines require the human account holder to perform the login, so even access the customer has authorized still violates those rules. These rules are designed to always make the bank in control of the user experience. Banking executives told Axios they fear the steps an agent might take once it holds that access, particularly if the user has granted it authority to move funds or complete purchases. In other instances, banks have tracked agents contacting call centers or initiating actions without clear direction from the account holder. One official described the situation as horrifying and said the industry holds grave concerns about agents gaining access and taking steps that could affect user finances. Morgan Stanley has begun blocking third-party agents from its platforms, according to people familiar with the decision. JPMorgan, Bank of America, and Citi are also restricting such agents. A Chase spokesperson said customers should not give their usernames and passwords to anyone. A Morgan Stanley spokesperson said the bank has not observed instances of rogue AI agents accessing its platforms. Many banks are authorizing or moving toward read-only access through open-banking connectors, the same type of limited access already granted to third-party budgeting tools. Even with that distinction, sources said banks broadly struggle to determine whether observed activity is user-directed or whether an agent has been authorized by the customer or has gained access on its own. The reports follow earlier statements from bank executives about external agents. On October 13, during JPMorgan’s third-quarter earnings call, CEO Jamie Dimon said the bank would not allow just any outside agent inside because it was too risky for clients and for the company, adding that “you don’t want rogue agents, and it could be thousands of them.” A bank spokesperson confirmed that third parties are not permitted to obtain customers’ credentials and log in on their behalf, while the bank continues to support secure, customer-authorized data sharing through open-banking providers. Read-only views of balances remain possible through those connectors; transfers and bill payments do not. Banks are investing in defenses against the agents and preparing “day after” plans that would let them secure or restore customer access if unauthorized activity occurs. Sources told Axios that cash on hand and experience with critical infrastructure give banks advantages in responding, yet the speed of the technology keeps revealing new weaknesses. Several executives said they recognize customers will want to use agents, but the risks of being too permissive remain too high for now. They want to block the bots mostly for the need to stay in control. But also a small part to keep the customer safe. Buckle up.