Overview
- U.S. District Judge Jed Rakoff in Manhattan, who ruled in February 2026, ordered defendant Bradley Heppner to turn over 31 documents he generated with Anthropic’s Claude.
- Rakoff said no attorney‑client relationship exists with a chatbot and noted Claude’s terms told users not to expect privacy in their inputs.
- Provider policies for OpenAI and Anthropic allow sharing user data with third parties, which undercuts any claim that AI prompts are confidential.
- Courts are split, as a Michigan magistrate judge treated a pro se litigant’s ChatGPT exchanges as her own work product and declined to compel production.
- More than a dozen law firms now urge clients to avoid consumer chatbots for legal matters, add contract warnings about privilege waiver, recommend closed enterprise AI, and suggest flagging when AI research occurs at counsel’s direction.