Judge says AI chats aren’t privileged
A U.S. federal judge ruled in a securities‑fraud case that conversations with AI chatbots are not protected by attorney‑client privilege, prompting lawyers to warn clients not to treat those chats as confidential. (reuters.com). Legal observers say prosecutors and civil litigants may now seek chatbot transcripts in discovery, changing how people and firms use generative‑AI for advice. (thedailyrecord.com) (republicworld.com).
A federal judge in New York said chats with an artificial intelligence bot are not protected by attorney-client privilege. (harvardlawreview.org) The ruling came in *United States v. Bradley Heppner*, a securities-fraud case in the Southern District of New York. Judge Jed Rakoff held on February 17, 2026, that 31 documents tied to Heppner’s exchanges with Anthropic’s Claude could be inspected by the government. (perkinscoie.com) (woodsrogers.com) Heppner was indicted in October 2025 and arrested in November 2025 after prosecutors accused him of fraud and false-statement offenses tied to his work as a public-company executive. Harvard Law Review’s summary says Federal Bureau of Investigation agents seized the AI-related files during a search of his home. (harvardlawreview.org) (perkinscoie.com) Attorney-client privilege usually protects confidential communications between a client and a licensed lawyer made to get legal advice. Rakoff said the Claude exchanges failed that test because Claude was not a lawyer and the chats were sent to a third-party platform. (foxrothschild.com) (harvardlawreview.org) The court also rejected a second shield, the work-product doctrine, which can protect materials prepared for litigation. Perkins Coie said the judge found that client-generated AI materials created without direction from counsel did not qualify. (perkinscoie.com) One detail running through the opinion was confidentiality. Harvard Law Review’s account says the court pointed to Anthropic’s policy stating it collected user inputs and outputs, used them to train Claude, and reserved the right to disclose data to third parties, including regulators. (harvardlawreview.org) That reasoning has widened the warning beyond one defendant and one bot. Reuters, republished by Republic World and AOL, reported that more than a dozen large United States law firms have told clients not to discuss active legal matters with chatbots such as Claude and ChatGPT. (republicworld.com) (aol.com) Alexandria Gutiérrez Swette of Kobre & Kim told Reuters clients should “proceed with caution.” The same report said some firms are now adding contract language warning that sharing a lawyer’s advice with a chatbot can waive privilege. (republicworld.com) The ruling lands as consumer AI products openly describe ways user content may be stored or reused. OpenAI says ChatGPT content for individual services may be used to improve models unless users opt out, while Anthropic’s Privacy Center separates consumer plans from business offerings with different data rules. (openai.com) (privacy.claude.com) The immediate lesson from *Heppner* is narrower than “all AI use is unprotected” and broader than one criminal case. A public chatbot can look less like a lawyer’s office and more like a third-party service, and courts may treat the transcript that way. (harvardlawreview.org) (bloomberglaw.com)
