Citibank, PNC and Citizens have rewritten their customer terms so that people who give third-party AI agents access to their bank accounts are responsible for whatever those agents do, including losses after an agent is hacked. The three banks follow JPMorgan Chase and Bank of America, which announced similar policies in October.
Citi updated its Online User Agreement this week. Under the new terms, if an AI tool leaks a customer's credentials or is compromised, the customer is responsible for what attackers then do with the account. Citi, the third-largest US commercial bank, said the update is meant to give customers greater clarity about the risks of third-party AI tools and to reinforce that customers who grant those tools access are responsible for activity conducted through them.
PNC, the eighth-largest US bank, changed its terms on Wednesday, November 18. Citizens emailed customers this week telling them they are also responsible for anything their AI assistant does.
Key facts
- 01Citi updated its Online User Agreement this week to make customers liable for third-party AI agent activity, even after a hack.
- 02PNC changed its terms on November 18; Citizens emailed customers a similar warning this week.
- 03JPMorgan Chase and Bank of America adopted similar policies in October.
- 04Wells Fargo's new Online Access Agreement, effective November 24, treats transactions by AI agents as authorized.
- 05Citi says AI agents it provides itself carry no additional liability for customers.
The banks say they will not reverse transactions made by automated assistants, whether the customer intended them or not. Citi's terms draw a line around its own products: AI agents that Citi provides carry no additional liability for customers.
The earlier moves set the template. In October, Chase began blocking outside agents such as Muse and Instinct from reaching accounts directly, and on October 27 Bank of America asked customers not to share banking passwords with an AI assistant. Wells Fargo's updated Online Access Agreement, which takes effect November 24, treats every transaction made by a customer's delegates, AI agents included, as authorized.
The tools at issue can be handed bank and brokerage credentials to analyze spending and investments, offer advice, pay bills and execute transactions. Anthropic's personal finance tool, launched earlier in November, takes a narrower approach: it is billed as read-only and cannot move, manage or invest a user's money.
The policy shift also has a competitive dimension. AI agents that can compare rates and move money could help consumers avoid fees and shift cash out of low-interest deposit accounts. Service providers such as Uber and DoorDash have pushed back on outside agents in similar ways.
Under Regulation E, consumer liability for unauthorized electronic transfers is capped at low amounts when the customer reports quickly. By defining agent actions as authorized by the customer, the new terms move those transactions outside that protection. No court or regulator has tested that reading, and no US agency, including the CFPB, had commented on the changes as of Wednesday.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.




