Security

AI agents and bank accounts: why big US banks are blocking them, and what to do

Big US banks are blocking AI agents from bank accounts that log in with customer passwords. What changed, why, and safer read-only options.

By AgentPlaybook Real Estate · Published Oct 22, 2026 · 5 min read

Key takeaways

Several large US banks, including JPMorgan, Bank of America, Citi, Wells Fargo and Morgan Stanley, are restricting AI agents that log in with a customer's own password, according to Axios, the FT and CNBC reporting.
Banks' stated concerns: they can't tell whether an agent is acting for you, who is liable for an agent-approved payment is unclear, and agents could move deposits out quickly.
Most banks are steering toward read-only access through approved data providers rather than letting agents act.
If you've given an agent your online banking password, you may be breaking your bank's rules; read-only connections and password managers built for agents are safer.
Keep money movement as a human-only step, and check your own bank's policy for business accounts, since the reports focus on consumer accounts.

AI agents and bank accounts are on a collision course. If you have been letting an agent sign in to your bank, the rules are changing fast. Several of the largest US banks are restricting AI agents that sign in using a customer's own username and password, Axios reported on October 22, 2026. The names include Morgan Stanley, JPMorgan, Bank of America and Wells Fargo. According to The Decoder's summary of the report, that kind of access breaks bank policies that require account holders to log in themselves (The Decoder). We could not load the original Axios story directly, so the details below come from outlets that cite it, plus earlier Financial Times and CNBC reporting.

For small business owners who are starting to hand bookkeeping and bill chasing to agents, this matters. Here is what changed and how to keep your agent useful without putting your accounts at risk.

What the banks are doing

The moves have stacked up over about two weeks:

JPMorgan Chase said on its third-quarter results call that it would block AI agents from client accounts, citing the risk of "rogue agents" that could number in the thousands. It is building its own protective tools and won't let outside agents operate in its systems until they're ready (The Paypers).
Bank of America told CNBC on October 14 that it is limiting agents' access to consumer deposits while it works out who is liable for payments an AI authorises. It also said a human should stay in the loop to approve money leaving an account (via The Paypers).
Citi said it doesn't allow third-party AI agents or assistants to log into customer accounts or make transactions on its digital banking platforms. Customers can still grant limited, read-only access to certain Open Banking providers (via The Paypers, citing the FT).
Morgan Stanley already blocks AI agents entirely, but stressed that it has not seen any "rogue AI agents" on its platforms (The Decoder).

Banks are not the only ones closing doors. The Verge reported earlier in October that JPMorgan Chase and Charles Schwab are among the businesses blocking agents, alongside sites like Delta and Yelp (The Verge).

Why banks are worried

They can't tell who is logging in

Most banks want agents limited to read-only access, because many of them can't tell whether an agent is acting on a user's behalf or on its own (The Decoder). Mother Jones, also citing Axios, reported that a senior bank official said banks have trouble telling agents apart, and noted that Meta's Muse connects to accounts through Plaid (Mother Jones).

Nobody has settled liability

If an agent approves a payment you didn't mean to make, who pays? Bank of America says it is keeping restrictions in place until that is clear, and regulators reportedly think the law may need clarifying (The Paypers).

Deposits could move quickly

There is also a business reason. A report by Apollo's chief economist, cited by The Paypers, warned that AI assistants could sweep household cash into accounts paying 3.3% to 5%, against a 0.1% national average on checking accounts. It said JPMorgan, Bank of America and Wells Fargo hold $1.6 trillion in non-interest-bearing deposits, and paying 3% on that would cost them $47 billion a year. Wall Street executives have reportedly raised the rise of AI agents with the Federal Reserve, which declined to comment (The Paypers).

The safer ways agents can see your money

Blocking an agent that types your password into a browser is hard to enforce, The Decoder notes, because tools like Computer Use or OpenClaw can drive a browser like a person would. But you don't need to rely on that. Safer patterns are emerging:

Read-only data connections. Citi still allows read-only sharing with approved Open Banking providers, and The Decoder notes OpenAI's US financial service in ChatGPT uses read-only access to financial data.
Credentials the agent never sees. On October 22, 1Password launched 1Password for Agents. An agent asks for a login, you approve it on your own device, and 1Password fills it in so the password is never sent to the agent. It names Muse, Claude Duet and ChatGPT's dots among the agents it works with (1Password).
Hard limits on money movement. Anthropic says Claude Duet will not move money between accounts or trade securities even if you approve it (Anthropic).

What this means for small businesses

The reports focus mainly on consumer accounts. Business banking rules can differ, so check your own bank's terms. In the meantime, a few sensible steps:

Stop sharing your online banking password with any agent. Several big banks now say this breaks their rules, and it puts your account at risk if the agent is tricked. Our article on prompt injection in email shows how that can happen.
Use read-only feeds for bookkeeping. Most agent money jobs, like categorising transactions, matching invoices or drafting a cash-flow summary, only need to read. Connect your accounting software (which already pulls bank feeds) rather than the bank itself.
Keep payments human. Let the agent prepare a payment run or draft an invoice reminder, then make the payment yourself in your bank's app.
Watch for blocked access. If an agent suddenly fails at your bank, that may be the new policy working, not a bug. Don't try to get around it.
Write down what each agent can touch. A one-page list of agents, accounts and permission levels makes it easy to revoke access when someone leaves or a tool changes.

The prompts below handle the read-only side of money work and run in any chat app. Use the AI time savings calculator to see which finance task is worth automating first.

Frequently asked questions

Can I give an AI agent access to my bank account?

Several major US banks now restrict agents that log in with your password, and Citi says it doesn't allow third-party agents to log in or transact. Read-only access through approved data providers is still possible at some banks.

Which banks are blocking AI agents?

Reports name JPMorgan Chase, Bank of America, Citi, Wells Fargo and Morgan Stanley. Morgan Stanley blocks agents entirely. Schwab has also been reported as blocking agents.

Why are banks blocking AI agents?

They say they can't reliably tell whether an agent is acting for the customer, liability for AI-approved payments isn't settled, and agents could shift deposits quickly.

Is read-only access safe for an AI agent?

It is much lower risk because the agent can view data but can't move money. It still exposes financial information, so only connect accounts the task needs.

Sources

How we report: every fact, number and quote above comes from the linked sources, which we read before publishing. Product details can change quickly during previews, so check the vendor page before you buy. Spotted an error? Tell us via the link on our home page.