Banks and AI agents: Citi, PNC, Wells Fargo make you liable
Citi and PNC changed their terms on Nov 18: if you give an AI agent your bank login, you pay for what it does, even if it's hacked. Bank-by-bank list.
Checked and updated 19 November 2026 at 15:21 UTC

Key points
- Citi and PNC updated their customer terms this week. If you let a third-party AI agent into your bank account, you are responsible for what it does, including if the AI tool is hacked or leaks your login.
- JPMorgan Chase and Bank of America took similar positions in October, and Wells Fargo's new terms from November 24 make you liable for an agent's actions unless you can prove you revoked its access.
- The safe rule for now is simple. Don't give any AI assistant your banking password, and use read-only connections if a budgeting tool needs to see your balance.
If you give an AI agent access to your bank account, the biggest US banks now say anything it does is on you, even if the AI tool gets hacked. Citi updated its Online User Agreement this week and PNC changed its terms on November 18, 2026, joining JPMorgan Chase, Bank of America and Wells Fargo, according to TechCrunch and American Banker.
This matters because AI agents (assistants that click, log in and buy things for you) are moving into money. Shopping agents like those behind ChatGPT's wallet, travel bookers and personal finance bots all want a way into your money. Banks are drawing a line: their own AI tools are fine, outside ones are your risk.
Bank by bank: what each one says
| Bank | What changed | What it means for you |
|---|---|---|
| Citi | Online User Agreement updated the week of Nov 16 | You are responsible for activity through any third-party AI agent you give access to, and for what hackers do if that tool leaks your login. Citi doesn't let outside agents log in or transact; read-only open banking access is still allowed. |
| PNC | Terms changed Nov 18 | Details not public yet. PNC's CEO said in October that no bank will let AI agents transact without a human in the loop. |
| Wells Fargo | New terms effective Nov 24 | You stay liable for an agent's activity, even after telling it to stop, unless you can show evidence you revoked its access. |
| JPMorgan Chase, Bank of America | Similar policies announced in October | TechCrunch says their policies are similar to Citi's; Reuters reported both block external agents such as Meta's Muse. |
| Morgan Stanley, Charles Schwab | Said in October they block external agents | Outside agents can't get into customer accounts. |
| Fifth Third | Statement to Banking Dive | AI agents are not allowed to start transactions in customer accounts. |
| Citizens | Email to customers this week | Warns that you are responsible for anything your AI assistant does. |
Citi's spokesperson put it plainly to TechCrunch: customers who grant access to third-party AI agents "are responsible for activities conducted through those agents".
Why banks are doing this
Two reasons come up in the reporting. The first is security: an agent that holds your password is one more place for it to leak, and banks don't want to cover losses from a tool they never vetted. Citi wants a "know your agent" check, similar to the identity checks for new customers, before it opens any door to outside agents. The second is competition: banks want customers to make money decisions inside their apps, with their own assistants. Citi is building its own agents on an internal platform called Arc, and its terms say agents Citi provides can be used without extra liability.
American Banker points out the open question in Wells Fargo's version: if you did revoke an agent's access and it acted anyway, it is not yet clear what happens.
What you should do
- Never give an AI assistant your bank username and password. That is exactly the case these terms leave you liable for.
- Use read-only connections for budgeting. If an app only needs to see balances and transactions, connect it through your bank's open banking or data-sharing screen, not by typing your login into the app.
- Stop agents at the payment step. If you use a shopping or travel agent, let it fill the cart and pay yourself. United Airlines took the same line for bookings; see United's ban on AI agents.
- Remove old access. Check your bank's connected apps list and revoke anything you no longer use. With Wells Fargo's terms, keep a screenshot or email showing you did.
- Read your own bank's terms. Smaller banks and credit unions may follow soon.
Tools built to be read-only are a different case. Anthropic's personal finance tool, launched earlier in November, says it can't move, manage or invest your money; TechCrunch does not say the bank rules apply to it.
New to the idea of agents? Start with what an AI agent actually is.
Frequently asked questions
Can I let ChatGPT or another AI agent use my bank account?
Most big US banks say no or put the risk on you. Citi says it doesn't let third-party AI agents log in or transact, and its new terms make you responsible for activity through any agent you give access to. Read-only open banking connections are the exception at Citi.
Who pays if my AI agent is hacked and drains my account?
Under Citi's updated terms, you do. TechCrunch reports the update says the customer is responsible for anything hackers do if an AI tool leaks credentials or is compromised. Wells Fargo will hold you liable unless you can show you revoked the agent's access.
Are banks' own AI assistants covered by the same rule?
No. Citi's terms say customers can use AI agents that Citi itself provides without worrying about additional liability. The rules target outside tools.
Sources
- More banks say customers using AI agents are on their own, even if they get hacked (TechCrunch, Nov 18, 2026)
- Wells Fargo wants customers to prove they've revoked AI agents (American Banker, Nov 2, 2026)
- Citi takes cautious approach to AI agent access (Banking Dive, Oct 23, 2026)
- Introducing AI agents, the next phase of Citi's artificial intelligence journey (Citi, Apr 30, 2026)
Related
What is ARTEX? The AI hacking tool the FBI warned about (Nov 2026)
ARTEX is an AI hacking agent framework. The FBI says attacks took under 3 hours; Amazon says its agents stole AI API keys and tried crypto deepfakes.
UC ChatGPT Edu deal: free ChatGPT for UC students from Dec 1
The University of California is paying OpenAI $1.2M a year so all 10 campuses get free ChatGPT Edu and Codex from Dec 1. What students get and the limits.
TikTok AI checker and Likeness Detection: how they work (2026)
TikTok's free AI checker tells anyone if a photo or video was made with TikTok AI, no account needed. Likeness Detection flags deepfakes of verified creators.