Recent decisions on discovery of generative AI usage, particularly prompts, searches, and outputs, reflects a judicial instinct to apply familiar doctrinal framework to new technology, rather than create AI-specific discovery rules. Two decisions issued one week apart, Warner v. Gilbarco, Inc. (E.D. Mich. Feb. 10, 2026) and United States v. Heppner (S.D.N.Y. Feb. 17, 2026), illustrate that approach. Read together, Warner and Heppner establish that discoverability of AI-related materials turns not on the use of AI itself, but on traditional questions of privilege, work product, and relevance. For construction claims, whether AI-assisted work is subject to discovery will be based on how it is used, not the fact that it is used at all.
There Is No “AI Rule”
Warner and Heppner emphasize that generative AI does not alter existing legal standards. Instead, Warner and Heppner applied existing doctrine to AI-assisted activities. This is critical: discovery disputes about AI prompts or outputs are analyzed the same way as any other document.
Heppner: AI Use as Non-Privileged, Discoverable Material
In Heppner, a criminal defendant independently used a public AI platform to develop legal theories. The court held that neither attorney-client privilege nor work product doctrine applied. There was no attorney-client relationship; the communications were with an AI system, not a lawyer. There was no reasonable expectation of confidentiality; the platform’s terms allowed data retention and potential disclosure. There was no attorney direction or strategy; the defendant created the materials independently, not as part of counsel’s litigation preparation. The fact that the defendant later shared the materials with an attorney did not retroactively cloak the materials in privilege.
From a discovery standpoint, Heppner signals that AI prompts and outputs may be fully discoverable where a party conducts self-generated research or analysis outside the attorney-client relationship.
Warner: AI-Assisted Work Product Is Protected
In Warner, the defendants sought discovery into a pro se plaintiff’s use of AI, including prompts and outputs. The court denied the motion, finding the work product doctrine applied: the materials were prepared in anticipation of litigation; using AI did not constitute purposeful disclosure to a third-party; and the requested discovery essentially sough mental impressions, improperly exposing litigation strategy. The court also characterized broad requests for AI usage as disproportionate and akin to a “fishing expedition.” The AI prompts and outputs in Warner were the functional equivalent to drafts or internal notes and were therefore shielded when they reflected litigation strategy or mental impressions.
Reconciling Warner and Heppner
Although the outcomes differ, Warner and Heppner are not inconsistent; both apply the same analytical framework but reach different results based on the presence (or absence) of counsel; expectations of confidentiality; and whether the materials reflect litigation strategy. Heppner addresses independent, non-confidential AI use, while Warner addresses litigation-directed, strategy-laden use.
Relevance to the Construction Industry
Warner’s and Heppner’s treatment of discovery of AI prompts and outputs have immediate consequences for construction litigation, where disputes often involve complex technical analyses, delay claims, and document-heavy discovery. Contractors and consultants increasingly use AI tools to summarize contracts, model delay scenarios or analyze critical path method (CPM) schedules. If done independently (as in Heppner), those prompts and outputs could be discoverable as factual or analytical work. However, when AI is used to draft claims, requests for information, or expert narratives under counsel direction in anticipation of litigation, Warner suggests those materials may qualify as protected work product.
In summary, courts are not treating AI as exceptional; instead, they are asking traditional questions about confidentiality, attorney involvement, and its litigation purpose. For construction industry participants, the lesson is straightforward: AI-assisted work will be discoverable or protected based on how it is used, not whether it is used at all.
This article is provided for informational purposes only—it does not constitute legal advice and does not create an attorney-client relationship between the firm and the reader. Readers should consult legal counsel before taking action relating to the subject matter of this article.
