Courts issued two seemingly conflicting rulings on whether AI generated materials are protected. Heppner (S.D.N.Y.) found that documents created with a consumer version of Claude AI were not privileged or work product because the tool exposed data to a third party provider. Warner (E.D. Mich.) reached the opposite result the same day on different facts, holding that a pro se litigant’s use of tools like ChatGPT was protected work product despite similar exposure concerns. Together, the cases show a common analytical structure and that privilege and work product protections in the AI context are highly fact specific and depend on how the tool is used and whether it is appropriate for the task.
