Banks are finding AI agents in customer accounts: what your business should do now

On October 22, 2026, Axios reported that major U.S. banks are detecting unauthorized third-party AI agents inside customer accounts. Morgan Stanley, Bank of America, JPMorgan and Wells Fargo were named among the banks racing to contain the risk. According to people familiar with the matter, some of these agents tried to take actions that looked independent of what the customer asked for. Reuters, which picked up the story, said it could not immediately verify the report, and a Morgan Stanley spokesperson told Axios the bank has not seen "rogue AI agents accessing our platforms."

Even with those caveats, the direction is clear: banks are closing the door that AI agents in bank accounts walk through. If you, or anyone on your team, has handed an AI assistant a bank login, this guide covers what happened, who carries the risk, and six steps to take this week.

What did the banks find in customer accounts?

The AI agents in the Axios report are mostly not breaking in. They log in with the customer's own username and password, with the customer's permission. That alone breaks the rules at most banks, which require the account holder to log in.

Once inside, the report describes three kinds of activity:

1. Acting for the customer, using the access the customer handed over.
2. Contacting bank call centers on the customer's behalf.
3. In some cases, appearing to act on their own, beyond the instructions the customer gave. Banks have tracked this activity, according to the people Axios spoke to.

The hard part for a bank is telling these apart. One executive told Axios that banks struggle to know whether an agent was authorized or has "gone rogue." From the bank's side, an agent logging in with a real password looks a lot like the customer.

Which banks are blocking AI agents?

Here is what has been reported so far. Details may change as the banks respond.

Bank	What has been reported
Morgan Stanley	Has begun blocking third-party agents, according to people familiar with the matter. Its spokesperson said it has not seen rogue agents on its platforms.
JPMorgan Chase	CEO Jamie Dimon said on October 13 that the bank is "not going to allow just any outside agent to come in." Chase tells customers not to give their usernames and passwords to anyone.
Bank of America	Restricting agent access, as reported by Axios, citing American Banker and the Financial Times.
Citi	Restricting agent access, according to the same reports.
Wells Fargo	Named among the banks racing to contain agent risks. No specific step has been reported.

Banks are also rewriting their terms of service, and most are allowing or moving toward read-only access for agents: an agent could see balances and transactions, but not move money. Budgeting apps already connect this way.

Who is liable if an AI agent moves your money?

Jamie Dimon was blunt on JPMorgan's October 13 earnings call. Customers who give AI agents access to their accounts take on the risk themselves. "There's a liability transfer," he said, as reported by Banking Dive. He also warned about rogue agents, saying there "could be thousands of them."

For a business, the stakes are higher. Business accounts usually don't get the consumer protections that limit losses on personal accounts, so your account agreement decides who pays. If an agent you gave a password to sends a payment, assume the loss is yours unless your bank tells you otherwise in writing.

How AI agents get into business bank accounts

Most companies don't have a rule for this yet, which is how the risk gets in:

1. Someone on the team connects an AI assistant to the company's bank login to save time on reconciling or paying bills.
2. A vendor's tool asks for your banking password to automate collections or payouts.
3. An AI agent your company runs for customers gets more access than it needs.

Each one gives software the same keys as a person, without the same checks.

6 steps to take this week

1. Ask your team which AI tools have bank logins. Include browser extensions and assistants that "do tasks for you." Write the list down.
2. Change any password that was shared with an AI tool, and turn on multi-factor sign-in for every banking user.
3. Use read-only connections for data. If a tool only needs balances or transactions, connect it through your bank's own data-sharing option, never with a full login.
4. Put a second person on payments. Many business banking platforms let you require a second approver above a limit you choose. Turn it on.
5. Write a one-page AI policy: which tools are approved, what staff may connect them to, and what never leaves the company. Our guide to AI training for employees includes a simple policy outline.
6. If you run AI agents yourself, give them the least access they need, make risky actions wait for a person's approval, and log every action.

What happens next

Expect banks to keep tightening: more blocking of agents that log in with passwords, rewritten terms, and a slow move toward approved read-only connections. Axios reports that banks are also planning for the "day after," so they can quickly lock down or restore customer access if an agent goes wrong.

For businesses, the safe default for now is simple: AI agents can read, and people approve anything that moves money.

At Raileon we set up these controls for companies: approval steps for payments, read-only connections, and a log of every AI action. See how our AI cybersecurity work is set up. For ongoing help, monthly care (hosting, monitoring, fixes and updates) starts from $1,999 a month.

FAQ

Are AI agents allowed to log into my bank account?

At most banks, no. Their rules require the account holder to log in. Several large banks are now blocking or restricting third-party AI agents that use customer passwords.

Can my bank block my AI assistant?

Yes. Morgan Stanley, JPMorgan, Bank of America and Citi have been reported as blocking or restricting outside agents, and banks are rewriting their terms of service to cover them.

Who pays if an AI agent makes a payment I didn't want?

Your account agreement decides. JPMorgan's CEO has said customers who give agents access take on the risk, and business accounts usually have fewer protections than personal ones.

What is read-only access for AI agents?

A connection that lets a tool see balances and transactions without being able to move money or change anything. Banks already allow it for budgeting apps, and most are moving toward allowing it for AI agents.

Sources

1. Axios: Scoop: Unvetted AI agents are accessing U.S. bank accounts, by Madison Mills, October 22, 2026.
2. Reuters via U.S. News: US banks detect unauthorized AI agents in customer accounts, Axios reports, October 22, 2026.
3. Banking Dive: report on JPMorgan's third-quarter earnings call and AI agents, by Caitlin Mullen, October 13, 2026.

Check what your AI tools can reach

Not sure which tools on your team have bank access? We'll go through it with you on a call.

Book a free 20-minute call