Citi, Citizens Shift AI Agent Liability to Customers
Citibank, N.A. has updated its Online User Agreement to state that customers who grant account access to third-party AI agents are responsible for activity conducted through those agents, even following a credential leak or compromise of the AI tool itself, TechCrunch reported on November 18, 2026 TechCrunch. A Citi spokesperson told the outlet the bank "recently updated" its terms "to provide greater clarity about those risks and to reinforce" customer responsibility for third-party agent activity.
Citizens Also Warns Customers
Separately, Citizens emailed customers during the week of November 16, 2026, warning they are responsible for actions taken by their AI assistants TechCrunch. The available reporting does not include the bank's full legal name, the complete text of the email, or an official public statement from Citizens confirming the policy beyond what TechCrunch described.
PNC Claim Not Yet Verified
Several outlets, including Unite.AI and AI Chat Daily, have repeated a claim that PNC Financial Services Group also changed its customer terms to impose similar liability. No primary PNC document, dated official notice, or verifiable quotation from PNC has been identified to confirm this. Readers should treat the PNC portion of the story as reported-but-unconfirmed pending a primary source.
What the Terms Cover
The reporting indicates the new or updated terms address transactions made by automated assistants, including unintended transactions, but the available sources do not establish that any of the banks have adopted a uniform, categorical refusal to reverse every such transaction TechCrunch. No court case, regulator statement, lawsuit, or penalty tied to these policy changes has surfaced in available reporting.
Why It Matters
The shift comes as more bank customers connect third-party AI agents to online banking to automate payments, transfers, and account management. By placing liability on the customer rather than the bank, the updated terms effectively treat an AI agent the same way banks have long treated a human authorized user: if it acts on your account, you own the consequences, hacked or not. With no regulator statement yet on record addressing this liability shift, the terms currently stand as unilateral bank policy rather than a settled legal framework.


