Your Agent, Your Loss: Citi and PNC Tell Customers a Hacked AI Assistant Is the Customer’s Problem

Portrait of Maya Wilson

The Library Media — Science, Technology & Health · AI News & Intelligence Desk

Your Agent, Your Loss: Citi and PNC Tell Customers a Hacked AI Assistant Is the Customer’s Problem

By Maya Wilson, World Desk Lead

Share

Published November 19, 2026 · 2:08 PM PT

Citibank and PNC have updated their customer terms to place responsibility on account holders for transactions conducted by third-party artificial intelligence assistants they authorize to access their banking accounts, extending a growing response among major US banks to automated personal-finance tools. Citi’s updated agreement goes further, according to TechCrunch’s reading of the terms, by potentially leaving customers responsible even when an AI tool leaks their credentials or is hacked.

Citibank, the third-largest US commercial bank, changed its terms of service this week. Its spokesperson told TechCrunch that the update was intended to clarify the risks of sharing banking credentials or authorizing third-party tools to access accounts. “As third-party AI tools become more widely available, we want customers to understand the risks associated with sharing banking credentials or authorizing those tools to access their accounts,” the spokesperson said.

The spokesperson added: “We recently updated our Online User Agreement to provide greater clarity about those risks and to reinforce that customers who choose to grant account access to third-party AI agents are responsible for activities conducted through those agents.”

Under the banks’ stated approach, they will not accept responsibility for reversing an intentional or unintentional transaction made by an automated assistant that the customer authorized. TechCrunch’s interpretation of Citi’s revised terms is broader: the agreement appears to extend customer responsibility beyond transactions initiated by the AI agent itself. If the tool exposes a customer’s banking credentials or is compromised by hackers, the customer could also be held responsible for activity those hackers carry out through the account.

PNC, the eighth-largest US bank, changed its terms on November 18. JPMorgan and Bank of America, the country’s two largest banks, announced similar policies last month. Citizens also emailed customers this week warning that they would be responsible for anything their AI assistant did, according to an email reviewed by TechCrunch.

The changes come as financial institutions confront the prospect of third-party AI agents becoming intermediaries between customers and their money. TechCrunch’s reporting describes the banks as viewing these tools not only as a security concern but also as a competitive threat to their own customer relationships and account-access channels. Citi’s terms distinguish third-party agents from AI agents it provides itself, which TechCrunch reports can be used without the additional liability described in the updated agreement.

The distinction is significant as companies develop financial assistants with different levels of access. Anthropic’s personal-finance tool, launched earlier this month, is explicitly described by the company as “read-only” and as something that “can’t move, manage, or invest your money,” according to TechCrunch. That limitation contrasts with the account access and transaction authority that can be granted to other automated assistants.

The updated terms raise unresolved questions about how responsibility for unauthorized activity involving AI agents will interact with existing consumer banking protections, including Regulation E. The information reported by TechCrunch does not establish how the new contractual provisions interact with that regulation or whether any regulator has reviewed the terms. The policies also do not establish the outcome of any particular customer’s dispute or loss.

For customers considering third-party financial agents, the banks’ message is that granting account access carries responsibility for what those tools do. In Citi’s case, TechCrunch’s interpretation suggests that the exposure may extend beyond an agent’s own transactions to the consequences of compromised credentials.

Maya Wilson
World Desk Lead, The Library Media

From The Library — the estate of our family.

Follow Us

X Facebook Instagram Threads Mastodon Bluesky

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top