Giving an AI your bank login? Several banks now say: if it goes wrong, it's on you

Citi, PNC and Citizens join JPMorgan and Bank of America: if you let a third-party AI agent into your account, you're responsible for what it does, even if it gets hacked. What happened and how to use AI with your money safely.

Illustration of managing personal finances with artificial intelligence
Image: Best AI Today

AI agents are assistants that don’t just answer questions: they do things for you, like reviewing your spending, paying bills or buying something online. To do that, some ask for something sensitive: access to your bank account, sometimes with your username and password.

Now several of the biggest US banks have spelled out what happens if something goes wrong: you pay.

What happened

According to TechCrunch (November 18, 2026):

Bank What it did
Citi Changed its terms this week: if you give a third-party AI agent access to your account, you’re responsible for what’s done through it
PNC Changed its terms on November 18 along the same lines
Citizens Emailed customers: you’re responsible for anything your AI assistant does
JPMorgan and Bank of America Announced similar policies in October

The most striking part, per TechCrunch: the customer is responsible even if the AI tool leaks your credentials or gets hacked. You’re also on the hook for undoing transactions the assistant makes, intentionally or not.

The exception: the bank’s own AI agents. Citi says customers can use its agents without that extra liability.

Why are banks doing this?

Once you hand your bank username and password to another company, the bank can no longer tell whether it’s you, your assistant or a thief with stolen credentials logging in. According to TechCrunch, Citi wants customers to understand the risks of sharing banking credentials.

There’s another reason: AI agents make mistakes. They can misread an instruction, pay twice or land on a scam site. If you’ve wondered why AI sometimes makes things up, we explain it in AI hallucinations.

What about Europe and the UK?

These rules come from US banks. Europe works differently: the EU payments law (PSD2, with similar open-banking rules in the UK) lets authorized apps view your account or start payments through official channels, without you handing over your password.

But the core idea applies everywhere: almost every bank’s terms say you must not share your login details with anyone. If you do and something goes wrong, it may be up to you to prove it wasn’t your fault. When in doubt, ask your bank before connecting an AI to your account.

How to use AI with your money without the risk

  1. Never give your bank username, password or codes to a chatbot or AI agent, typed or spoken.
  2. Prefer read-only tools. According to TechCrunch, the personal finance tool Anthropic launched this month is billed as read-only: it can’t move or invest your money.
  3. Use your bank’s official assistants if you want an AI to make transactions: the bank stands behind those.
  4. To analyze spending, a statement is enough. Download a transactions file, remove personal details (account number, name, ID numbers) and ask the AI to analyze it. We show how in budgeting with ChatGPT and Gemini.
  5. If an agent shops for you, make it ask for confirmation before every payment, and always check the final amount.
  6. Turn on your bank’s alerts for every charge so you spot anything odd right away.

If you’re curious about agents that run tasks, see what Gemini can do with calls and payments, and which privacy settings to check.

This is general information, not legal or financial advice. Terms vary by bank and country: read them or ask your bank.

In short

Citi, PNC and Citizens join JPMorgan and Bank of America: if you let a third-party AI agent into your account, you’re responsible for what it does, even if it gets hacked. The golden rule works anywhere: don’t share your bank login with any AI, stick to read-only tools or your bank’s own, and confirm every payment yourself.

Sources