Several of the largest banks in the United States are cracking down on AI agents that get into customer accounts by using login details handed over by the account holders. According to a report from Axios, the banks include Morgan Stanley, JPMorgan, Bank of America and Wells Fargo, all of which are limiting access by outside AI agents.
The banks say the practice breaks their rules, which require account holders to sign in themselves. In some cases, the agents seem to operate independently rather than following direct commands from the user, which adds to the concern.
Morgan Stanley already blocks AI agents completely, and the report says JPMorgan, Bank of America and Citi are taking the same approach. Morgan Stanley emphasized that it has not detected any "rogue AI agents" on its platforms.
For most banks, the preferred middle ground is read-only access. Under that model, an agent could look at account data but could not take any action on the account. The banks acknowledge that customers want to use AI agents, but they argue the security risks are currently too great to allow more.
A core difficulty is identification. Many banks cannot tell whether an agent is working on a customer's behalf or reaching into an account on its own, which makes it hard to judge whether activity is legitimate.
Enforcing a ban may prove difficult. Customers can give agents access to their accounts through a web browser, using tools such as Computer Use or open-source agent software like OpenClaw. Because the agent works through the same interface a person would, stopping it from the provider side is a challenge.
Meanwhile, the technology industry is moving toward finance. OpenAI has introduced a dedicated ChatGPT service for financial tasks in the United States, and it is limited to read-only access to financial data, matching the level of access the banks themselves say they are most comfortable with.
