The list of major US banks telling customers that third-party AI agents are their problem keeps growing. Citi, the third-largest commercial bank in the United States, has updated its Online User Agreement to make clear that customers who grant account access to AI agents are responsible for what those agents do — even if the tool leaks their credentials or gets hacked and attackers drain the account. PNC, the eighth-largest bank, changed its terms the same week, and Citizens emailed customers warning that they are liable for their AI assistants' actions.
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, a Citi spokesperson said in a statement reported by TechCrunch. 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.
The two largest banks got there first. JPMorgan and Bank of America announced similar policies last month, and the pattern is now consistent across the top of the US banking market: hand your credentials to an external AI assistant, and intentional or accidental transactions, plus the fallout from any compromise of the tool itself, land on the customer. Notably, the liability carve-out does not extend to the banks' own products — Citi's terms say customers can use the bank's own AI agents without additional liability, drawing a sharp line between first-party and third-party automation.
The banks are reacting to a fast-moving change in what consumer AI can do. AI agents can now be given credentials to analyze spending, offer advice, make transactions and pay bills: Google's Tasks agent completes purchases through Google Pay, OpenAI launched a ChatGPT Wallet that stores cards and lets its agent finish checkouts, and Anthropic launched a personal finance feature in Claude earlier this month — explicitly billed as read-only, and unable to move, manage or invest money.
Liability is only half of the story. Banks also see agents as a competitive threat: software that helps customers dodge fees or shift money out of low-interest deposit accounts attacks revenue streams that fund bank margins. That is why the industry posture looks less like customer protection and more like gatekeeping — the same institutions blocking agent traffic on one front are defining the terms under which it is allowed on another.
For consumers the practical takeaway is uncomfortable: the convenience of an agent that manages your money comes with a liability model built for a world where a human typed the password. Regulators have yet to say whether such terms will hold up, but until they do, the largest banks are converging on the same answer — your AI agent, your problem.
Comments (0)
Log in to join the discussion
Log InNo comments yet