United States v. Heppner

The defendant was indicted in the Southern District of New York on securities-fraud, wire-fraud, conspiracy, and records-falsification charges. Before his arrest, and while under investigation, Heppner used Claude to run queries and generate documents relating to the government’s investigation and potential defenses. Some queries incorporated information he had learned from his defense attorneys. During a search of his residence, federal agents seized devices containing these AI-generated transcripts and documents. He asserted that the records were protected by attorney-client privilege and the work-product doctrine because they were prepared to facilitate discussions with his legal team and later shared with counsel. The Department of Justice moved for a ruling that the AI documents were not privileged and could be reviewed by the prosecution team.

The court granted the government’s motion, holding that the defendant failed to meet his burden under both attorney-client privilege and the work-product doctrine. A commercial AI chatbot cannot form an attorney-client relationship and is legally akin to consulting a search engine or a nonlawyer friend. The court also found no reasonable expectation of confidentiality under the public platform’s terms and privacy policy, which permitted logging queries, using inputs for model training, and disclosing data to authorities on request. Forwarding a preexisting nonprivileged interaction to counsel did not retroactively cloak it with privilege. The materials also were not work product because the defendant generated them on his own initiative, rather than at counsel’s direction or under counsel’s supervision.