Major US banks restrict AI agents that log in with customer credentials
Several major US banks are restricting external AI agents that get into customer accounts using credentials the customers hand over, according to Axios, as relayed by The Decoder. The banks named are Morgan Stanley, JPMorgan, Bank of America and Wells Fargo. The agents are getting in without any approval from the bank, and that breaks bank policies requiring account holders to log in personally. The Decoder adds that some AI agents appear to act on their own, without direct instructions from their users.
The response goes beyond restriction at some institutions. Morgan Stanley already blocks AI agents entirely, and JPMorgan, Bank of America and Citi are doing the same. Morgan Stanley also stressed that it has not seen any "rogue AI agents" on its platforms.
Most banks would like to limit agents to read-only access: an agent could 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. A further problem is detection. Many banks can't tell whether an agent is acting on a user's behalf or accessing an account on its own.
The route in is simple. Users can give their agents access to accounts through a browser, using tools like Computer Use or open-source agent software such as OpenClaw. The Decoder's view is that blocking this on the provider side will be hard. Meanwhile, OpenAI has launched a dedicated ChatGPT service for financial tasks in the US with read-only access to financial data.
Key facts
- Axios reports that Morgan Stanley, JPMorgan, Bank of America and Wells Fargo are restricting external AI agent access to customer accounts.
- The access in question uses customer-provided credentials without bank approval, which violates policies requiring account holders to log in personally.
- Morgan Stanley already blocks AI agents entirely; JPMorgan, Bank of America and Citi are doing the same. Morgan Stanley says it has seen no "rogue AI agents" on its platforms.
- Most banks want to limit agents to read-only access (viewing data, no actions), and many cannot tell whether an agent acts for a user or on its own.
- OpenAI has launched a ChatGPT service for financial tasks in the US with read-only access to financial data.
Why it matters
Banks are drawing a line on who may sit at the keyboard of a customer account. Their rule is that the account holder logs in personally, and an agent that borrows a customer's credentials breaks it. The story shows a clash between customers who want to hand tasks to AI agents and banks that say the security risks are too high for now. The proposed middle ground is read-only access, where an agent can see account data but take no actions. OpenAI's new US ChatGPT service for financial tasks, which also has read-only access to financial data, follows that same model.
Who it affects
Customers who give an AI agent their bank login are the first in line, since the access violates bank policy and is being restricted or blocked. Morgan Stanley, JPMorgan, Bank of America and Wells Fargo are named as restricting external agent access, and Citi is named alongside the banks that block agents. AI agent makers and users of browser-based tools such as Computer Use or the open-source OpenClaw are affected too, as are providers like OpenAI that want to offer financial services.
How to use it
The source gives no instructions, but its practical picture is clear. Banks are willing to consider read-only access, where an agent views account data without carrying out actions. Handing an agent your credentials so it can log in through a browser is the practice banks say violates their policies. OpenAI's financial ChatGPT service is described as read-only.
How solid is it
This is secondhand. The Decoder relays Axios, and the details of the original Axios report are not reproduced beyond what The Decoder states. The bank list is stated as "including", so it is not necessarily complete. The statement that Morgan Stanley has seen no "rogue AI agents" comes from the bank itself. The claim that some agents appear to act on their own is The Decoder's hedged wording, and the judgement that blocking will be hard is The Decoder's too, not a bank statement.
Risks and caveats
No date is given for when the banks introduced the restrictions or when Axios published its report, and no figures such as accounts affected, incidents or losses appear. No specific incident of an AI agent causing harm is described. Whether Wells Fargo blocks agents entirely is not stated; only Morgan Stanley, JPMorgan, Bank of America and Citi are said to block. Detection is a stated weakness: many banks can't tell whether an agent is acting on a user's behalf or on its own, and The Decoder expects blocking to be hard because users can grant access through a browser.
“Morgan Stanley stressed that it has not seen any "rogue AI agents" on its platforms.”
— The Decoder, relaying Morgan Stanley's position