Major US banks are blocking AI agents that sign in using customer credentials

Background: AI Agents Reach Bank Accounts

AI agents are software tools that can perform tasks for users, and they are now moving into sensitive banking environments. According to Axios, several major US banks, including Morgan Stanley, JPMorgan, Bank of America and Wells Fargo, are restricting external AI agent access. The issue is not simply that users are experimenting with new tools, but that those tools can use customer-provided credentials to reach bank accounts without bank approval.

That creates a direct conflict with bank policies requiring account holders to log in personally. The banks’ assumption is that access to an account should be tied to a human customer at the moment of entry. AI agents, however, may operate with credentials supplied in advance, and some appear to act on their own, without direct instructions from users.

Development: Banks Move From Concern to Blocking

The response has hardened as the risk has become clearer. Five banks named in the report are now part of the tightening: Morgan Stanley, JPMorgan, Bank of America, Wells Fargo and Citi. Morgan Stanley already blocks AI agents entirely, and JPMorgan, Bank of America and Citi are doing the same. Wells Fargo is also restricting external AI agent access.

The banks are not describing a confirmed wave of malicious activity. Morgan Stanley stressed that it has not seen any “rogue AI agents” on its platforms. The concern is broader: the banks cannot reliably tell whether an agent is acting on a user’s behalf or accessing an account on its own.

Most banks want to limit agents to read-only access, letting them view account data but not carry out any actions. The banks say customers want to use AI agents, but the security risks are too high for now. The practical effect is that banks are choosing restriction first, even while customer demand points in the other direction.

Latest Progress: Read-Only Services Face Provider-Side Gaps

Users can give their agents access to accounts through a browser, using tools like Computer Use or open-source agent software such as OpenClaw. Blocking that on the provider side will be hard, because the access route is broad and can be initiated by the customer rather than by the bank.

At the same time, the industry is testing a safer shape. OpenAI has even launched a dedicated ChatGPT service for financial tasks in the US with read-only access to financial data. That matches the banks’ preferred direction: show data, do not act. It suggests that read-only access may become the first accepted form of agent-assisted banking.

Current Stage: Early Containment, Not Full Adoption

The sequence so far is clear: AI agents begin using user credentials to reach bank accounts; banks respond by restricting or blocking external access; read-only services emerge as a possible compromise. The current stage is early containment. Banks are not rejecting customer interest in AI agents, but they are treating external access as too risky until they can distinguish user-directed sessions from autonomous activity.