A week before S.D.N.Y. Judge Jed Rakoff’s unfortunate rejection of both privilege and work product protection for AI-related communications in United States v. Heppner, 820 F. Supp. 3d 292 (S.D.N.Y. 2026), another federal court found the work product doctrine for AI-related communications applicable in Warner v. Gilbarco, Inc., 820 F. Supp 3d 629, 636 (E. D. Mich. 2026). That more expansive view is consistent with the broad array of those who can create protected work product in anticipation of litigation.
Unfortunately, a judge in another AI-related matter missed the mark. In Shealy v. Seaside Investments, LLC, No. 2684CV00799-BLS2, 2026 Mass. Super. LEXIS 135 (Mass. Super. Ct. June 16, 2026), plaintiff “directed” his romantic partner to use AI in preparing a response letter to a creditor’s default notice. The court rejected plaintiff’s work product claim for his partner’s prompts and the AI output – warning that considering the partner “to be a party’s representative in these circumstances would . . . expand the work product doctrine without meaningful limitation.” Id. at *4. But just a few sentences earlier, the court quoted its work product rule – which on its face protects litigation-motivated documents prepared “by or for” a party or the party’s “representative . . . or agent.” Id. at *2 (emphases added). The plaintiff’s partner was clearly acting for a party and presumably would also satisfy the “agent” requirement.
As courts puzzle through the more difficult AI privilege implications (presumably focusing on the danger of strangers’ access to AI prompts and outputs), they should have an easier time relying on their own expansive work product language when the users or their agents (or those assisting them) rely on AI.