Which banks block AI agents? The full list (2026)
Which banks block AI agents: Chase, Citi, Bank of America, Wells Fargo and Schwab, what each has said, and the read-only route banks say they still allow.
Five of the biggest US banks have now said, in public or in their customer terms, that an AI agent can't simply log in with your password: JPMorgan Chase, Citi, Bank of America, Wells Fargo and Charles Schwab. What none of them has ruled out is read-only access that you approve through your bank's own sign-in, and that is the door Era uses.
This page keeps the list in one place, bank by bank, with what each one said and what still works for you.
Which banks block AI agents?
Chase, Citi, Wells Fargo and Schwab all block agents that sign in with a customer's own username and password, and Bank of America tells customers they are responsible for anything done with shared credentials. Chase says outright that read-only access through open-banking providers is allowed.
The block is aimed at the password, and your balances and transactions can still reach an assistant by another route.
| Bank | What it has said | What it means for you |
|---|---|---|
| JPMorgan Chase | On 13 October 2026 a spokeswoman said "we do not permit third parties to access accounts by obtaining customers' banking credentials", while read-only access through open-banking providers is allowed. Its updated Digital Services Agreement takes effect on 1 January 2027 and holds customers responsible for an agent's actions. | Read-only access through an approved link works. A password in an assistant does not. See Chase's AI agent terms. |
| Citi | On 16 October 2026 Citi said it does not permit "third-party AI agents or assistants to log into a customer's account or transact directly within our digital banking platforms." | Same line: no logins by agents. |
| Bank of America | A customer email of 30 October 2026 said that if you share your credentials, "you're responsible for activity resulting from that access." | The risk of a shared password sits with you. |
| Wells Fargo | Its Online Access Agreement takes effect on 24 November 2026: instructions from an AI agent count as yours, you must revoke an agent's authority yourself, and the bank may block AI tools. | Read Wells Fargo's AI agent terms before you hand an agent anything. |
| Charles Schwab | Its chief executive, Rick Wurster, said Schwab is blocking unauthorized agents, including Meta's Muse, and asked "who's liable?" Schwab has also said it will let clients connect their own assistants to its own MCP servers. | Schwab is closing the password door and promising a front door of its own. See is Muse safe with your bank login. |
| RBC, TD and BMO | The big Canadian banks have taken the same line since 28 October 2026. | Canadian readers will find the local picture on our Canadian pages as they publish. |
Does Chase block ChatGPT?
Chase blocks ChatGPT, Claude or any other assistant from signing in to your account with your Chase username and password. It does not block you from sharing read-only data through an open-banking provider that you approve. So ChatGPT can't be handed your Chase login, though it can see your Chase balances and transactions through an authorised link.
That distinction is worth understanding because it decides which of your options work. A password gives an assistant everything the Chase website can do, which is why banks object. A read-only link you approve in Chase's own flow gives it a view of your accounts and nothing else, and you can take it back whenever you like.
Will other banks follow?
Probably, and the direction is clear from the wording already in use. Banks are writing the same two ideas into their terms: an agent that uses your credentials is you, and you answer for what it does. Each new set of terms moves that way.
The pattern is likely to spread to more banks over the coming months.
For readers, that settles a practical question early. Build your setup on a route a bank has no reason to object to, since a connection that depends on a password is one terms update away from breaking.
What should I do if my bank is on the list?
Take your password out of any assistant, then link your accounts through a read-only connection and let the assistant read from there. That is the whole plan, and it is the one the banks' own statements point to. You keep using ChatGPT, Claude, Gemini or Meta's Muse, and none of them ever holds your bank login.
This is what Era does. You link your bank once, signing in inside your bank's own flow through a regulated data network, and Era holds read access that you can revoke. Every assistant you use connects to Era, so you approve the bank once and not once per assistant. Era connects accounts across the US, the UK, Canada, France, Germany, Spain, and 40+ countries in all.
How do I connect my bank to ChatGPT or Claude without a password?
Link your bank in Era, then connect your assistant to Era. The assistant asks Era for your balances, spending, financial health and its analysis, holdings or documents, and Era answers from the accounts you linked. The setup takes a few minutes and works the same for every assistant.
The step-by-step guides are here:
- Connect Claude to your bank account
- Connect ChatGPT to your finances
- Connect any AI agent to your bank accounts
Once connected, the assistants share one memory: tell Claude your savings goal on Monday and ChatGPT knows it on Tuesday. You can also bring in statements, payslips and tax forms through the Document Library, share a space with a partner or housemates, and move to a different assistant later without redoing a thing.
Is it safe?
Yes, when the route is read-only and revocable, and the password never leaves the bank. Era never stores your bank login, never sells your data and never uses it to train models, and you can cut any assistant off from Era in one step.
The longer answer is in is it safe to connect your bank to ChatGPT or Claude.
Related reading
Last updated