Overview
- Attorneys across the country are issuing fresh advisories telling clients that chats with tools like ChatGPT and Claude can be demanded by prosecutors or opponents in lawsuits.
- Guidance accelerated after Judge Jed Rakoff’s February ruling that ordered Bradley Heppner to produce 31 Claude-generated documents in his New York fraud case.
- Rakoff said an AI platform cannot have an attorney‑client relationship with a user and noted Claude’s terms telling users not to expect privacy in their inputs.
- OpenAI and Anthropic say in their policies that they may share user data with third parties, which undercuts claims that chats are confidential and can trigger a waiver of privilege.
- Firms now recommend closed enterprise AI, lawyer‑directed prompts with explicit language, and contract warnings, with Sher Tremonte and Debevoise offering model guardrails while a Michigan ruling treating some chats as personal work product shows courts remain split.