Big US Banks Restrict AI Agent Access to Customer Accounts

Morgan Stanley, JPMorgan Chase, Bank of America and Citi now fully block AI agents from logging into customer accounts. Wells Fargo takes a different approach: it limits them to view-only access, Axios reports. The agents in question log into a bank account using the customer's own password, often through tools like Computer Use (a feature that lets an AI agent control a web browser the way a person would) or OpenClaw (an open source AI agent program). Some of these agents act without a direct command from the user. Banks say they often cannot tell who is actually running a session, a person or an agent. That is why most of them want to limit agents to viewing account data rather than letting them take action.

This points to a bigger problem for AI agents working with real accounts. Hand an agent a full password, and it gets the same powers as the account owner. There is no way to limit what it can do after that. Narrow, task-specific access is much harder to abuse than a master password. More and more companies are letting agents into real accounts. This gap will likely matter more over time, not less.