Five Major U.S. Banks Are Blocking Outside AI Agents From Accounts

Five major U.S. banks are blocking or restricting outside AI agents that log into customer accounts with the customer’s own credentials, and the open-banking rule that would have given those agents a regulated route in is on hold. The reporting dates to October 22 and relies on people familiar with the matter rather than official bank disclosures.

Which banks are blocking AI agents

At least five banks are named as blocking external agents: JPMorgan Chase, Citi, Bank of America, Wells Fargo and Morgan Stanley. Yahoo Finance and American Banker both cite the five-bank count.

The story started with Axios. According to a Yahoo Finance summary, Axios reported on October 22 that banks were detecting outside agents operating inside customer accounts. Some of those agents took actions that seemed independent of what users had asked for, citing people familiar with the matter.

Reuters published a parallel account the same day. Citing two people familiar with the situation, it reported that major banks were blocking third-party agents from operating inside accounts and making transactions. Gulf News also covered the blocks.

The list is not limited to the five. Charles Schwab’s CEO said the firm was blocking unauthorized agents that try to move client money, including Meta’s Muse. That detail comes from a Financial Times report carried by Yahoo Finance. The exact date of the FT report could not be confirmed, so its timing should be treated as approximate.

What the banks have said on the record

Jamie Dimon, JPMorgan Chase’s chief executive, said the bank would limit AI agents’ access because “you don’t want rogue agents and there could be thousands of them,” according to the Financial Times. The word “thousands” signals the scale JPMorgan expects, even though it is a warning rather than a count.

Citi was more specific about its rules. In a statement to Retail Banker International, the bank said, “we do not permit third-party AI agents or assistants to log into a customer’s account.” Citi does allow certain read-only access through approved open-banking providers, according to the reporting.

Morgan Stanley’s position is the least intuitive. A spokesperson told Axios the bank had not observed “rogue AI agents accessing our platforms,” yet it began blocking third-party agents anyway as a precaution. The bank has not said whether those blocks respond to any confirmed incident.

The table below summarizes each named firm’s position as it appears in the reporting.

Why banks say they cannot tell an agent from the customer

The core technical problem is verification. Banks can confirm which credentials were used to open a session, but they cannot confirm the human intent behind it, according to Yahoo Finance’s five-bank report. A correct password tells the bank nothing about whether the customer asked the agent to move that money.

The concern, as reported, is less about data theft than about banks being unable to tell whether an agent is doing what the customer asked. That is a different problem from credential theft, and a correct password cannot solve it.

Nothing in the reporting establishes how many accounts were affected or any verified customer losses. An October 30 AI news roundup reported that no money was stolen in the cases it described. Those details matter, because the blocks rest on what people familiar with the matter described, not on disclosed incident counts.

The FT report adds two more threads. Regulators have reportedly privately supported the banks’ blocking decisions, and Wall Street executives have contacted Federal Reserve officials about AI agents used for transactions. Neither point appears in an official public statement in the reporting reviewed.

American Banker raises the obvious weakness. Banks may not be able to block every agent, and if one bank refuses them, agents could open accounts at less restrictive firms. That makes industry-wide blocks hard to hold.

The open-banking rule that would have offered a path stalls

The open-banking rule meant to replace password-based access is on hold. Without it, no regulated route exists for AI agents into bank accounts, and the blocks fill that gap with a flat refusal instead of a shared standard. Artificial Intelligence News connects the blocks directly to the stalled rule.

Former White House AI adviser David Sacks has argued that banks should support user-directed agents with grant, limit and revoke controls, with read-only access as an interim step. An open standard led by Sierra and Meta would let personal AI agents identify themselves and limit their access to what a customer authorized. Those are the building blocks a revived rule would need, but the reporting does not say banks have adopted either approach.

The idea of one shared signal is familiar to readers of USA 24’s coverage of a single robots.txt opt-out, which OpenAI, Google, Microsoft and Anthropic have backed. That case involves training data rather than bank logins, but the logic is similar. Both sides need a machine-readable answer instead of case-by-case negotiation.

What it means for consumers and small businesses

If you use an AI budgeting or money-management app, some features may stop working, especially those that log in with your bank password, according to Gadget Review. Tools that connect through a controlled data pathway such as Plaid may fall within the read-only access banks are moving to allow. Terms vary, so read the connection details instead of assuming.

Small business owners face a narrower but more immediate issue. Anyone using AI agents for payments or bookkeeping should expect more friction, and should confirm with the bank before automating transfers. American Banker reports the same pressure on business accounts.

Consumer use of agents is still early. USA 24 reported that DoorDash’s CEO said AI agents had placed only ‘several hundred’ orders. Banks are setting their rules before that volume grows, which is a reason to watch how they treat read-only access now.

How to protect your bank account from AI agents

The practical steps are simple, and none requires waiting for a rule.

• Check how your AI finance tool connects to your bank: a password login, an aggregator or a read-only link.
• Ask your bank whether it allows third-party AI agents to act on your accounts.
• Do not give any AI assistant your bank password.
• Start any agent on read-only access, which lets it view statements without moving money.
• Revoke access for tools you no longer use, and confirm with your bank before automating any transfer.

What to watch next

Several questions remain open. The reporting does not say how many accounts were affected or what the agents attempted. It is unclear whether Morgan Stanley and other precautionary blocks answer confirmed incidents or simply reflect risk. Accountability is also unclear when an agent acts beyond instructions, and no date has been given for reviving the open-banking rule.

The next useful signal will likely come from the banks themselves, through earnings calls, customer notices or a disclosed incident count. Watch whether any of the five publish a formal policy for read-only access, and whether the open standard gains bank backers. If the blocks harden into industry policy, people who rely on AI money tools will have to weigh convenience against a connection their bank will accept.