AI chats can appear in court
U.S. lawyers are warning that AI chatbot conversations may be admissible in legal proceedings after a recent ruling highlighted that chats could be used against users. The warning applies to both consumer and corporate use of AI assistants and underscores legal exposure tied to internal or external AI interactions. (reuters.com)
A federal judge in New York has ruled that some chatbot conversations can be used in court instead of being treated like private talks with a lawyer. (usnews.com) The ruling came in February in *United States v. Heppner*, a criminal case before Judge Jed S. Rakoff in the Southern District of New York. Rakoff said documents a defendant created with Anthropic’s Claude were not protected by attorney-client privilege or the work-product doctrine. (orrick.com) Bradley Heppner, the former chief executive of Beneficient, used Claude on his own after learning he was under investigation, then shared about 31 AI-generated documents with his lawyers. Federal agents later seized those documents during a search of his home. (dlapiper.com) Attorney-client privilege is the rule that usually shields confidential communications between a client and a licensed lawyer. Rakoff said a public chatbot is not a lawyer, and that alone defeated Heppner’s privilege claim. (crowell.com) The judge also said confidentiality was missing because the material was created with a third-party, publicly accessible system. Several law firms have since warned clients that pasting legal advice or strategy into a chatbot can waive privilege that would otherwise protect it. (orrick.com) The warning is not limited to criminal defendants. Lawyers told Reuters that civil litigants, companies and employees who use tools like ChatGPT, Claude or Gemini for sensitive legal or regulatory questions could face discovery fights over those chats. (usnews.com) Some firms are now putting the risk into engagement letters. Reuters reported that New York litigation firm Sher Tremonte recently told clients that sharing lawyer communications with a chatbot could destroy attorney-client privilege. (usnews.com) The case does not mean every AI-related document is automatically discoverable. DLA Piper said the decision turned on narrow facts: a non-lawyer used a public tool, on his own initiative, after the platform disclaimed privacy, and only later sent the material to counsel. (dlapiper.com) Still, the practical advice from law firms has been blunt: treat anything typed into a consumer chatbot as material that could be subpoenaed, seized or handed over under legal process. OpenAI and Anthropic both say they can disclose user information in response to valid legal requests. (openai.com) (privacy.claude.com) Heppner has pleaded not guilty to fraud-related charges tied to an alleged scheme involving more than $150 million, and his privilege fight has already changed how U.S. lawyers talk to clients about artificial intelligence. (justice.gov)
