Overview
- Judge Jed S. Rakoff in United States v. Heppner held that a defendant’s chats with the public AI platform Claude were neither attorney‑client privileged nor protected work product because they were not confidential and were not created at counsel’s direction.
- Investigators seized about 31 documents reflecting the AI interactions, and the court noted that sending those materials to lawyers after the fact did not create privilege, with platform disclaimers of legal advice further undercutting protection.
- Analyses emphasize that courts are applying traditional privilege and work‑product tests to GenAI data by scrutinizing who used the tool, for what purpose, and under what confidentiality terms.
- Other recent rulings point in different directions, including a Michigan magistrate’s order rejecting discovery into a plaintiff’s AI use and declining to find waiver, and decisions recognizing protection for counsel‑directed AI testing or litigation preparation in closed environments.
- Practitioners are updating policies to steer work through counsel, favor enterprise AI with non‑retention terms, document AI use, and address GenAI in ESI protocols and Rule 502(d) orders to reduce waiver risk.