Particle.news

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

Banks are shifting liability to customers with new terms that treat third‑party AI‑agent actions as customer‑authorized.

Overview

  • Citibank, which updated its online user agreement on Wednesday, Nov. 18, now says customers are responsible for any actions taken by third‑party AI agents, including when an agent leaks credentials or is compromised.
  • JPMorgan Chase, Bank of America, PNC, Wells Fargo and Citizens have moved to tighten contract language, block unvetted agents and steer users toward read‑only, bank‑approved connectors or regulated open‑banking flows.
  • Federal supervisors including the CFPB and the OCC have warned banks they cannot erase consumer fraud protections by contract, but regulators have not yet clarified how those rules apply when customers authorize AI agents to act.
  • Customers who share login details or give agents full account access face a higher risk of being denied reversals for mistaken or hijacked transfers and may bear the financial losses under the new terms.
  • The industry shift follows October detections of AI agents logging into accounts and has pushed some AI firms to market read‑only finance tools while prompting banks to adopt tighter technical controls and vetting for connectors.