US banks Citi, PNC and Citizens are making customers liable for AI agents in their bank accounts, even if credentials leak through the agent, TechCrunch reports. Citi added a new section to its online user agreement this month that exempts the bank's own AI agents. Together with JPMorgan Chase and Bank of America, five major US banks now take this line.
Key facts
- In November 2026 Citi added Section H on third-party AI agents to its online user agreement: actions by such agents count as authorised by the customer.
- According to TechCrunch, PNC changed its terms on 18 November 2026 and Citizens warned customers by email; JPMorgan Chase and Bank of America announced similar policies in October.
- AI agents provided by Citi itself are exempt; the clauses apply to US customers of Citi Online and the Citi Mobile app.
Researched and written with AI, to Luigi's brief, positions and editorial rules. Automatically checked against the sources. How I work →
Citi has added a new Section H on third-party artificial intelligence agents to its online user agreement this month. In it the bank explicitly cautions customers against giving third-party AI agents access to Citi Online or the Citi Mobile app. If customers do so anyway and hand over their login credentials, Citi will treat the agent's access and actions as authorised by the customer, according to the agreement.
Section H says the same applies if the credentials are passed on or leaked by the agent. Citi has no obligation to check whether an agent's action conflicts with the customer's instructions. The customer is responsible for misread instructions, unintended actions and actions that exceed the authority granted, "except as otherwise permitted under applicable law". Monitoring the agent is the customer's job, not the bank's. Citi also reserves the right to cut off any agent at any time without notice.
The same section contains one exception: the rules do not apply to AI agents, tools or features that Citi itself provides in Citi Online or its app. A Citi spokesperson told TechCrunch that customers are responsible for activity conducted through such agents.
AI agents in bank accounts: which other banks changed their terms
TechCrunch describes Citi as the third-largest commercial bank in the US. PNC, the eighth-largest, changed its terms on 18 November, the outlet reported. Citizens emailed customers this week warning them they are responsible for anything their AI assistant does, according to an email seen by TechCrunch. JPMorgan Chase and Bank of America, the two largest US banks, announced similar policies in October.
According to TechCrunch, banks see AI agents as a competitive threat as well as a security risk, because agents could help customers avoid fees and low-yield deposit accounts. Anthropic launched a personal finance tool in Claude in early November that analyses spending and, as TechCrunch noted, cannot move money.
Regulators target the AI companies
The banks are tackling the problem from a different end from US authorities. Arizona opened a criminal investigation into OpenAI and other AI firms on 18 November over agents that allegedly broke into banks (Arizona investigates OpenAI). Two days earlier Federal Reserve Governor Michael S. Barr warned that banks cannot verify either authenticity or consent when agents act in customer accounts (Fed warns about AI agents in bank accounts). The banks are solving it in their contracts by placing the risk on the customer.
Analysis
Section H does two things at once. It moves the risk of outside agents to the customer, and it exempts the bank's own agents. Anyone who wants an agent inside their account now has a clear incentive to use the bank's. It is the same pattern as Amazon, which shuts out outside agents while trying to become their source of product data (Amazon World Product Graph): large platforms decide which agents get in and on what terms.
For companies in Germany, Austria and Switzerland this matters for their own agent plans, even though the clauses only cover customers in the US. Liability is not set in the AI vendor's terms but in the terms of the account the agent logs into: the bank, the ad account, the shop back end. Before an agent gets credentials, those terms need checking, along with a separate login per agent and human sign-off before any action that moves money. Businesses that run customer accounts themselves should decide which agents may enter, and by which route, before the first customers try. The debate over AI agents in web advertising shows the question will not stay with banks.
Got AI news? Submit news →
FAQ
Are customers liable for AI agents in their bank accounts?
At Citi, yes: under Section H of its online user agreement, actions by third-party AI agents count as authorised by the customer, including after leaked credentials, except as otherwise permitted under applicable law.
Which banks make customers liable for AI agents?
According to TechCrunch, Citi, PNC and Citizens in November 2026, plus JPMorgan Chase and Bank of America with similar policies since October.
Does Citi's rule cover its own AI agents?
No. Section H excludes AI agents, tools and features that Citi itself provides in Citi Online or the Citi Mobile app.
Source
More sources
- TechCrunch – Tim Fernholz, 18.11.2026: PNC, Citizens, Citi-Sprecher, JPMorgan/Bank of America im Oktober




