Osegi
Business·2 min read

Citi AI agents policy: you pay if your AI tool gets hacked

Citi now says customers who give banking logins to third-party AI agents are responsible for what happens next, including if the tool is hacked.

A person looking at a banking app on a phone while an AI assistant icon hovers over the account screen.
Image: TechCrunch

If you hand your bank login to an AI agent and something goes wrong, Citi says that is your problem. The bank, the third largest commercial bank in the US, updated its Online User Agreement this week to say so in plain terms.

The idea behind these tools is simple. You give an AI agent the credentials to your bank account and other financial portals, and it analyzes your spending, offers advice and can even handle transactions and bill payments. Citi's new terms say customers who grant that access are responsible for activity carried out through the agents. A spokesperson said the update was meant to give "greater clarity" about the risks of sharing banking credentials.

Citi goes a step further than some rivals. If the AI tool leaks your credentials or is hacked, the customer is responsible for whatever the hackers do with the account. Citizens emailed its customers this week with a similar warning, and PNC, the eighth largest US bank, changed its terms today. JPMorgan and Bank of America announced similar policies last month.

Why are banks pushing back on AI agents?

Partly it is security. Banks do not like unreliable automated actors inside their systems, and they do not want to be on the hook for reversing a transaction an agent made, intentional or not.

Partly it is competition. Markets briefly wobbled at the idea of AI tools helping people dodge unnecessary fees or low-interest deposit accounts that fatten bank margins. If agents become the main way people deal with their money, banks want to control that route. Citi's terms note that customers can use the AI agents the bank itself provides without worrying about extra liability.

What does it mean for AI finance tools?

Trust is already a hurdle. Many people are wary of letting an AI read their email or chats, and bank accounts raise the stakes. Anthropic's personal finance tool, launched earlier this month, is billed as "read-only" and says it "can't move, manage, or invest your money."

The pattern is familiar: AI labs promise agents will change how people use services, then the services say hold on a minute. It has happened with car-hailing and food service too.

Apps worth a look

SUMO: the number puzzle

Tap the right tiles, add them up, hit the target. Easy to learn, hard to master. Free, for all ages.

Tags:citiai agentsbankingpersonal financejpmorgancybersecurity

More from this day

See all →

More in: Business

See all →