AI chats may be discoverable
U.S. lawyers are warning clients that chatbot conversations might not be protected by attorney-client privilege after a federal ruling indicated AI chats can be discoverable in litigation. Reporters note a judge explicitly found chatbot conversations unprotected in a securities‑fraud case, prompting firms to rethink confidential use of AI tools. (reuters.com) (thedailyrecord.com)
A federal judge in New York has ruled that a defendant’s chatbot conversations were not protected by attorney-client privilege and could be turned over in court. (harvardlawreview.org) The case is *United States v. Bradley Heppner* in the Southern District of New York. Judge Jed S. Rakoff issued the written decision on February 17, 2026, after prosecutors sought access to about 31 documents reflecting Heppner’s exchanges with Anthropic’s Claude. (perkinscoie.com) (harvardlawreview.org) Heppner had been indicted in October 2025 and arrested in November 2025 on fraud-related charges tied to his time as a public company executive. According to court summaries, he used Claude after receiving a grand jury subpoena and after learning he was a target of the investigation. (perkinscoie.com) (harvardlawreview.org) Attorney-client privilege is the rule that usually shields confidential communications between a client and a lawyer made for legal advice. Rakoff said the Claude exchanges failed that test because Claude was not a lawyer and because the chats were made through a third-party platform. (natlawreview.com) (harvardlawreview.org) Rakoff also pointed to Anthropic’s privacy terms, as described in court analyses, which said the company collected user inputs and outputs, used them to train Claude, and reserved the right to disclose data to third parties, including government authorities. The court said that undercut any claim that the exchanges were intended to stay confidential. (harvardlawreview.org) (perkinscoie.com) The ruling did not say every use of artificial intelligence destroys privilege in every setting. Harvard Law Review’s analysis said the opinion was framed as a “question of first impression nationwide,” while also arguing that a more fact-specific approach could leave room for privilege in some lawyer-directed uses of artificial intelligence. (harvardlawreview.org) Law firms reacted by warning clients not to paste sensitive facts, defense theories, or draft responses into consumer chatbots and assume those chats are private. Legal alerts published after the ruling said using a public-facing tool as a stand-in for a confidential lawyer conversation can create discovery risk. (natlawreview.com) (perkinscoie.com) The decision landed as courts and lawmakers were already tightening scrutiny of artificial intelligence in the justice system. Bloomberg Law reported in April 2025 that 39 federal judges had issued standing orders on artificial intelligence use in their courtrooms, and Politico reported on March 19, 2026 that lawmakers were preparing a bipartisan bill to study artificial intelligence in federal courts. (news.bloomberglaw.com) (politico.com) For anyone using a chatbot to think through a lawsuit, subpoena, or criminal case, the Heppner ruling draws a sharp line: a chat window is not a lawyer’s office. (harvardlawreview.org)
