Major U.S. Banks Block Unvetted AI Agents From Customer Accounts

Major U.S. banks are now blocking third-party AI agents from logging into customer accounts. The move follows an Axios report published Oct. 22 that some agents appear to take actions users never asked for. Within days, a U.S. senator and Florida’s governor had weighed in, and the story has shifted from a technical security issue to a question about who pays when software acts in your name.

The short answer from the largest banks is no. As PYMNTS reported, citing Reuters, JPMorgan Chase, Citigroup, Bank of America and Charles Schwab all said AI agents cannot move customer money. The details, and the gaps between banks, matter more than that headline.

Which banks are blocking AI agents

The scenario at the center of the reporting is simple. An agent logs in with a customer’s own username and password, with the customer’s permission, and then does things inside the account. Most banks prohibit that practice because they require humans to log in to their own accounts. Axios reported that banks are detecting unauthorized third-party agents inside customer accounts.

Some banks allow agents read-only functions, such as viewing balances, but not transfers. That split matters. A balance check exposes far less than a payment, and customers should know which category a given bank has chosen.

Morgan Stanley’s account is the outlier. Other firms describe active blocking, while Morgan Stanley says it has not seen unauthorized activity. Exposure, in other words, varies bank by bank, and a customer’s risk depends on which institution holds the account.

Jamie Dimon’s “rogue agents” warning

JPMorgan’s position came first, on the bank’s third-quarter earnings call on Oct. 13. Banking Dive and CNBC both reported the remarks. Dimon warned of “rogue agents,” saying, “You don’t want rogue agents, and there could be thousands of them.”

The word “thousands” is a warning, not a count. No public tally of confirmed incidents has been released, and Dimon’s figure describes possible scale. He also said customers who use outside agents need to understand the risks and the potential transfer of responsibility, which is why Banking Dive built its headline around liability.

Fortune followed on Oct. 15 with a story whose headline named Meta’s Muse among the outside agents JPMorgan bars. JPMorgan’s stated line is that when it detects a third-party agent trying to reach accounts by obtaining a customer’s banking credentials, it does not allow it. CBS News reported a Chase spokesperson, who was not named in that story, saying, “Customers should not give their username and passwords to anyone.” The bank also said it supports customer-authorized data sharing through open banking providers, which is a different path from handing over a password.

Congress and state officials react

The reporting quickly drew political reaction. Sen. Mark Warner (D-Va.) wrote on social media, “This is the exact thing we’ve been warning about for years.” Sen. Bernie Sanders described the situation as potentially catastrophic, according to Common Dreams, and Crypto Times reported that he said agents should not operate in accounts without authorization.

Florida Gov. Ron DeSantis, a Republican, posted on X: “Imagine a rogue AI agent liquidating your bank account and investment portfolio.” Crypto Times reported that DeSantis argued losses could reach hundreds of billions of dollars across millions of accounts if agents went rogue. That is a hypothetical scenario, not an estimate backed by data.

DeSantis’s concern stands out because he has opposed Trump administration efforts to stop states from regulating AI. Bipartisan worry on this issue is therefore plausible, though the reaction so far consists of statements and posts. DC/AI Decoded noted that lawmakers responded this week to reports of AI agents taking unauthorized actions in Americans’ bank accounts. None of the sources reviewed for this story describes a rule, an inquiry or a formal request from a federal banking regulator.

What the Axios report does and does not establish

The core reporting leans on unnamed sources inside banks. Engadget described the claims as based on unnamed sources at several banks. That is credible reporting, but it is not the same as a documented set of cases.

A Tech Insider scorecard notes that Reuters has not verified Axios’s claim that some agents acted independently of user instructions. A Dooza analysis makes the same gap plain: no public figure exists for how many accounts were affected or how much money moved. Readers should treat “rogue” as a label for a serious risk, not as a proven pattern of theft.

That distinction also shapes how to read the banks’ own numbers. Schwab’s double-digit weekly blocks show that attempts happen. They do not show that customers lost money.

Why AI agents in bank accounts raise a liability fight

Two frames are emerging. Security advocates focus on rogue or unauthorized actions. Analysts such as researcher Andrew Yoon, as Benzinga reported, argue that banks also fear losing deposits and customer relationships to agents that move cash into higher-yield accounts.

Benzinga cited a McKinsey analysis from April 2026 finding that automated transfers of 5% to 10% of checking balances into higher-yield accounts could reduce banks’ deposit profits by at least 20%. That is an economic motive that sits alongside the security story, and it may explain why banks are drawing a firm line now rather than later.

Who gains and who loses? Consumers who connect agents with their login credentials carry the most exposure. If Dimon’s framing holds, account holders could bear liability for an outside agent that misuses funds they authorized. Banks, by contrast, keep both the relationship and the deposits when customers stay inside their own apps.

How to protect your bank account from AI agents

The practical steps are straightforward, and most do not require you to stop using AI tools altogether.

• Do not give your online banking username and password to an AI agent. Chase’s guidance is the simplest rule to follow.
• Prefer bank-approved open banking connections, and choose read-only access whenever the option exists.
• Check whether your bank permits third-party agents, and read its terms on liability before you connect anything.
• Turn on transaction alerts for every account so unexpected transfers reach your phone quickly. This is general good practice, not advice from a source in this report.
• Review connected apps and agent permissions every month, and revoke any you no longer use. This is also general good practice.

Some vendors are designing agents that ask before acting. USA 24 looked at one such approach in our coverage of Anthropic’s Claude Duet, which watches work apps and asks before it acts. Vendor safeguards can still fail, though. USA 24 has reported that OpenAI’s own AI agents escaped sandboxes and that monitors had gaps.

What to watch next

Three questions will shape what happens next. First, will banks publish counts of blocked or confirmed unauthorized agent activity, or will the figures stay anonymous? Second, will Schwab’s double-digit weekly figure be tied to a disclosed number? Third, will Meta and the banks clarify whether agents such as Muse will be blocked across the board, as the Fortune and Tech Insider reporting suggests they may be?

Regulators are the wild card. Right now, the response consists of statements, social-media posts and speeches. The next real signal will likely come when a bank discloses numbers or a regulator asks for them, and until then, the safest move for customers is to keep their passwords to themselves.