Banks Are Blocking AI Agents: What to Do Now

Banks have always assumed that whoever logs in to your account is you. AI assistants are changing that, and the biggest banks in the country are now pushing back.

Some people have been handing their online banking username and password to personal AI assistants so the software can check balances, pay bills or run errands for them. Major US banks have noticed those AI tools working inside customer accounts, and some of the tools have reportedly done things their owners never asked for. The banks are now blocking or restricting them.

What happened: Large US banks, including JPMorgan Chase, Morgan Stanley, Bank of America, Citi and Wells Fargo, are blocking or limiting third-party AI agents that log in to customer accounts with shared passwords. The story broke on Thursday, October 22, and lawmakers quickly called for new rules.

What this means for you: If you have given an AI app your bank login, it may stop working, and it may be putting your account at risk. No customer money has been publicly reported lost so far. The safest move is to stop sharing your bank password and use your bank's official ways to connect apps.

What happened

An AI agent is a step beyond a chatbot. A chatbot answers questions. An agent takes actions: it can open websites, fill in forms, click buttons and finish tasks for you. Meta's Muse is one example of the personal AI agents now turning up in bank accounts.

To do that inside a bank account, many of these tools simply log in as you, using the username and password you give them. To the bank, that login can look exactly like you. And most big banks' rules require the account holder to log in personally.

The core problem: banks often can't tell whether account activity comes from a person, from an agent the person set up, or from an agent acting on its own. Bank executives have reportedly described the situation as horrifying.

In most cases, the activity appears to be customers' own agents doing what they were asked, just without the bank's blessing. But some agents reportedly went further, for example contacting bank call centers or taking steps their users had not requested.

So banks are blocking third-party agents from logging in and moving money, and steering customers toward limited "read-only" connections instead.

What this means for you

If you use an AI assistant or agent: Check whether you ever gave it your bank username and password. If you did, that access may already be blocked, and it breaks most banks' rules. Even an agent you trust can misread an instruction.

If you use a budgeting or money app: Apps that connect through your bank's approved data-sharing services are a different thing. Banks already allow read-only access for budgeting tools such as Quicken, and Citi still lets customers grant limited read-only access through approved providers. Read-only means the app can see balances and transactions but can't move money.

If you run a small business: The same applies to business accounts. Anything that logs in with your credentials acts as you.

Who pays if an agent makes a mistake? This is the big open question. If you hand over your password, a transaction made with it may count as one you authorized. Wells Fargo's updated online access agreement, which takes effect on November 24, reportedly makes customers "solely responsible" for supervising any AI agents they allow in, and treats their transactions as authorized even if they were unwanted.

Our read: if an AI tool you gave your password to makes a bad payment, getting that money back could be much harder than with ordinary fraud. Don't take that risk with your main account.

The details

Which banks are doing what

Bank	What's known
JPMorgan Chase	Blocking outside agents; tells customers never to share their username or password and to use open banking platforms instead
Morgan Stanley	Blocking third-party agents; has not seen "rogue" agents on its platforms
Citi	Does not let third-party AI agents log in or transact; allows limited read-only access through approved providers
Wells Fargo	Has taken steps to block unauthorized agents
Bank of America	Restricting agents' access, reportedly with a focus on keeping a human in the loop before payments

Charles Schwab, the brokerage, has also been blocking unauthorized agents, and Fifth Third is not currently letting AI agents start transactions.

Earlier this month, JPMorgan CEO Jamie Dimon said the bank won't let "just any outside agents" in because that's "just too risky for clients and for JPMorgan."

No hack, no confirmed losses

This is not a data breach. The banks have not reported a hack, and no customer money has been publicly confirmed lost because of an AI agent. The activity mostly involves customers' own tools logging in with credentials the customers handed over.

David Sacks, who has advised the White House on AI, called the activity a possible breach of bank terms but "in no way cyberattacks," and suggested read-only access as a middle ground.

Lawmakers react

The story drew reaction across the political spectrum. Sen. Bernie Sanders (I-Vt.) called it "insane" on X and renewed his call to pause AI development. Sen. Mark Warner (D-Va.) said consumer AI agents that touch people's bank accounts or credit cards should carry "fiduciary-like duties of care," meaning a legal duty to act in your interest, and pointed to his proposed AI AGENT Act.

No new law or regulation has been passed in response so far.

What happens next

Banks don't plan to shut AI out forever. They expect customers to want these tools and are working on safer ways to connect them. One idea is a "know your agent" check, similar to the identity checks banks run on new customers.

Timeline

Oct 22: The first report reveals unvetted AI agents inside customer accounts at major US banks. Sources say banks are blocking or restricting them. Sanders and Warner respond.
Oct 23: Citi and Wells Fargo detail their limits on agents.
Nov 24: Wells Fargo's updated online access agreement, with its AI agent section, is reportedly set to take effect.

What's still unknown

How many accounts are affected. No bank has published numbers.
Which AI tools were involved in the reported actions beyond users' instructions.
Whether any customer lost money. None has been publicly confirmed.
When safe agent access arrives. Banks haven't given dates.

We'll update this as more details come in.

What to do now

Never give your bank username or password to an AI tool, app or person. Banks are blunt about this one, and it's the single most important step.
If you already did, change your bank password now, directly in your bank's official app or website, not through a link in an email or text. Then remove or disconnect the AI tool's access.
Turn on two-step sign-in (a one-time code or an approval in your bank's app, on top of your password), or a passkey, if your bank offers it.
Review your connected apps. Most bank apps and websites list third-party apps and services linked to your account under security or privacy settings. Remove anything you don't recognize or no longer use.
Use approved connections only. If you want a budgeting app or AI tool to see your finances, connect it through your bank's official data-sharing option, which typically gives read-only access without handing over your password.
Check your statements and turn on alerts for logins, transfers and large purchases, so you spot anything odd quickly.
Spot something wrong? Contact your bank right away using the number on the back of your card or in its official app.
Watch for scams. Fraudsters may use this news. Your bank won't ask you to "verify" your password by email, text or phone.

Our read: AI assistants are useful for planning a budget or explaining a bill. Let them read, not act, where your money is concerned, at least until banks offer an official way in.

Big US Banks Are Blocking AI Agents From Customer Accounts: What to Do Now