Morgan Stanley, JPMorgan block AI agents using customer logins

The standoff is not over whether customers want AI agents — banks concede they do — but over a gap in what a bank can verify. Account policies require holders to log in personally, and an agent presenting customer credentials does not meet that test. Even where banks might tolerate automation, many cannot distinguish an agent acting on a user's behalf from one accessing an account on its own, and some agents appear to act without direct instructions. That uncertainty is what pushes institutions toward outright blocks rather than case-by-case approval.

Morgan Stanley says it has not seen any "rogue AI agents" on its platforms, which suggests the restrictions are precautionary rather than a response to a specific incident. Meanwhile the tools that enable this access are spreading through ordinary browsers: Computer Use and open-source agent software such as OpenClaw let users hand over credentials without bank involvement. Blocking on the provider side will be hard, according to the report.

The emerging middle ground is read-only access. OpenAI has already launched a dedicated ChatGPT service for financial tasks in the US with read-only access to financial data, a model that gives customers visibility without letting an agent move money. Whether banks can enforce that line against browser-based agents they do not control is the open question.