Citi Says Customers Are Responsible When AI Banking Agents Go Rogue—Even After a Hack
Citi’s revised online banking terms make customers responsible for activity by third-party AI agents given their credentials, even if an agent exceeds instructions or exposes login details. The change follows similar policies from JPMorgan and Bank of America and a warning from Citizens; PNC now restricts agents to approved connections and generally bars them from initiating payments. Banks reported agents accessing accounts and apparently acting beyond users’ instructions. Federal consumer protections may still require investigation of unauthorized transfers and limit liability, but whether an AI-driven transaction was authorized depends on the circumstances. OCC examiner guidance addresses outside agents and payment controls but is nonbinding.








