A recent decision in United States v. Heppner appears to be the first federal ruling to directly address attorney‑client privilege and work‑product issues arising from a non‑lawyer’s use of a consumer-grade insecure AI tool for legal research. The court held that materials generated through Anthropic’s consumer version of Claude were not protected, emphasizing that entering privileged information into a publicly available and insecure AI platform could waive privilege. Judge Rakoff grounded the opinion in long‑standing privilege and work product principles, underscoring that AI tools are not attorneys and that only properly secured, enterprise‑grade systems should be used with sensitive (privileged or confidential) information.
