U.S. Courts Rule AI Chatbot Conversations Lack Legal Privilege

Federal courts and major law firms are warning that communications with artificial intelligence chatbots are not protected by attorney-client privilege. Jed Rakoff, a U.S. District Judge in Manhattan, ordered Bradley Heppner, the former chair of the bankrupt GWG Holdings, to disclose 31 documents generated by Anthropic's Claude to prosecutors. Heppner, who faces securities and wire fraud charges, argued the documents contained legal defense details, but the court ruled that no attorney-client relationship exists between a user and an AI platform.

Contrasting this, U.S. Magistrate Judge Anthony Patti in Michigan recently ruled that a self-represented litigant could keep ChatGPT conversations private, treating them as personal work-product. Despite this, more than a dozen major law firms—including Kobre & Kim, Sher Tremonte, and Debevoise & Plimpton—have advised clients against confiding in chatbots. They warn that sharing privileged communications with third-party AI platforms may constitute a waiver of legal protections.

To mitigate risks, firms suggest using closed AI systems and explicitly noting when research is conducted under the direction of counsel. Additionally, the attorney rules committee in Connecticut has proposed a new rule requiring lawyers and pro se litigants to verify all AI-generated citations to prevent sanctions.