Overview
- U.S. District Judge Jed S. Rakoff held that a defendant’s exchanges with Anthropic’s public Claude platform are not covered by attorney-client privilege or the work-product doctrine.
- The court cited the platform’s privacy policy allowing data retention and disclosure and emphasized that the AI use was initiated by the client, not at counsel’s direction.
- Agents had seized roughly 31 AI-related memoranda from Bradley Heppner’s devices, and the court found that sharing those materials with lawyers after the fact did not create protection.
- The decision framed work product as protecting attorneys’ mental impressions, rejected client-initiated AI materials as qualifying, and disagreed with more expansive readings in earlier SDNY rulings.
- Other courts have reached different results in fact-specific settings, including a Michigan decision treating GenAI as a tool and denying discovery of AI-use materials, and a prior case protecting counsel-directed prompts as opinion work product.