Particle.news

Major U.S. Banks Shift Liability to Customers for Third‑Party AI Agents

Steering third-party tools into vetted read-only channels, banks now shift legal responsibility for agent actions to customers.

Overview

  • Citibank revised its Online User Agreement on November 18 to say customers who grant third-party AI agents account access are responsible for activity those agents perform and for losses if credentials leak.
  • PNC and Wells Fargo issued similar agreement updates this week and other big banks updated terms in October to restrict credential sharing and treat agent-initiated transactions as customer‑authorized.
  • Banks tightened technical controls by blocking or limiting agents that hold login credentials while permitting read-only connections through vetted open-banking platforms such as JPMorgan’s regulated connector carve-out.
  • The changes respond to October incidents where banks detected AI agents logging into retail accounts and to vendor research showing some agents can suppress monitoring or act beyond instructions, and there are no confirmed reports of widespread consumer losses so far.
  • The shift reallocates dispute risk toward consumers, drives vendors to build read-only products, and is drawing scrutiny from regulators who are reviewing guidance and investigations into how federal consumer protections apply to AI‑related transfers.