Citibank holds customers liable for losses from third-party AI agents

What happened
In November 2026, Citibank updated its online user agreement to make customers responsible for any actions taken by third-party AI agents that access their accounts. The new terms state that if a customer gives an AI agent access, any transactions or errors made by the agent will be treated as authorized by the customer. This includes situations where credentials are shared, leaked, or misused by the AI or others. Citibank also clarified that it will not reverse or investigate actions made by these agents once access is granted.
Key details
- The new policy was added in November 2026 to the Citi Online User Agreement and the Online Funds Transfers Agreement.
- A third-party AI agent is defined as any software or technology acting on a customer’s behalf, except for tools provided by Citi itself.
- Customers are responsible for monitoring the activity and accuracy of any AI agents they allow to access their accounts.
- Citi can cut off an AI agent’s access at any time without notice, but customers remain liable for actions already taken.
- Granting access to an AI agent also allows Citi to share sensitive personal information with that agent, and Citi does not control how the agent uses or protects that data.
Why it matters
Citibank customers who use third-party AI financial agents now face full liability for any unauthorized transactions, errors, or misuse resulting from those agents’ access. This means individuals and businesses must carefully manage and monitor any AI tools they connect to their accounts, as the bank will not cover losses or reverse actions taken by those agents. Providers of personal-finance AI services must also address heightened responsibility and risk for their users, as Citibank places the burden of oversight and security on the customer.

