As lawyers’ AI use dramatically expands, courts have begun to address both types of evidentiary protections for AI-related communications and results. Popular public AI service providers’ ugly disclaimers of confidentiality presumably jeopardize attorney-client privilege protection for lawyers’ input or results. What about the hardier work product doctrine protection?
In United States v. Heppner, — F. Supp. 3d —, 2026 U.S. Dist. LEXIS 32697 (S.D.N.Y. Feb. 17, 2026), highly regarded S.D.N.Y. Judge Jed S. Rakoff rejected privilege protection after pointing to warnings that users could not expect confidentiality for their inputs or for the AI’s outputs. But surprisingly, Judge Rakoff also rejected work product protection despite: (1) the pertinent criminal rule’s extension of that protection to materials “made by the defendant”; and (2) acknowledging that “the work product doctrine may apply to materials generated by non-lawyers.” 2026 U.S. Dist. LEXIS 32697, at *11, *13.
Presumably lawyers can assure privilege protection by using one of the increasing number of vendors promising confidentiality for their communications and results. On the work product side, a Detroit magistrate judge’s opinion issued exactly a week before Judge Rakoff’s opinion helpfully applied the work product rule as it is written. Next week’s Privilege Point describes that case.