Two Federal Courts, Divergent Paths: What Heppner And Warner Mean For Artificial Intelligence
On our February 23rd McV Alert (Federal Court in New York Issues Landmark Opinion on AI, Privilege, and Client‑Generated Legal Research), we highlighted an important ruling from the Southern District of New York in United States v. Heppner,[1] where the court held that a criminal defendant’s self‑directed legal research using Anthropic Claude was not protected by attorney–client privilege or the work‑product doctrine.
On that same day, however, Magistrate Judge Anthony Patti from the Eastern District of Michigan reached a different conclusion about AI‑related materials in Warner v. Gilbarco, Inc.[2]
In a pro se employment discrimination case, the plaintiff used ChatGPT to assist in the drafting of litigation documents. The court denied the defendant’s motion to compel production, holding that the AI‑assisted materials were protected as work product prepared in anticipation of litigation under Rule 26(b)(3). Judge Patti stressed that using an AI tool does not, by itself, constitute disclosure to a third party in a manner that risks adversary access—absent a showing that the material was shared with, or was likely to reach, the opposing party. The opinion characterized generative AI systems as tools, not persons, and declined to treat the plaintiff’s reliance on ChatGPT as a waiver of protection simply because the platform is publicly accessible. Notably, the decision stands in contrast to Heppner’s emphasis on platform privacy terms but does not undermine the broader doctrinal principles at play.
Viewed together, Heppner and Warner do not announce competing legal theories. Instead, they illustrate that privilege outcomes hinge on specific factual circumstances. Four factors appear central to the analysis: ( 1) how the AI tool is used; (2) which platform is used and whether it is consumer‑grade or enterprise‑secured; (3) who directs the AI use (counsel versus client or pro se litigant); and, (4) what the AI platform’s terms of service permit regarding data handling, training, and confidentiality. The emerging bottom line is that privilege and work‑product protections are stronger when AI is woven into counsel‑directed workflows inside a confidential, contractually restricted, enterprise‑grade environment. Conversely, they are weaker when clients independently use public, unsecured AI systems outside the supervision of counsel. Both cases reaffirm that the attorney-client privilege ultimately tied to the human professional relationship. AI platforms are tools, not privileged intermediaries.
For clients and counsel, the lessons stemming from these cases point towards governance, instead of avoidance. Effective privilege‑preserving AI practices require secure platforms, contractual guarantees of encryption, disciplined content‑handling protocols, and robust access controls—conditions that ensure AI tools operate within the protected perimeter of the attorney–client relationship.
As courts continue to confront AI‑related privilege issues—and more cases will undoubtedly emerge—the early jurisprudence underscores a consistent theme: facts matter, human oversight matters, and the legal profession must structure AI usage to preserve the core protections on which the attorney—client relationship is based and the client’s trust depends. In short, as the doctrine evolves, the lawyer remains the essential safeguard of attorney-client privilege.
[1] United States v. Heppner, No. 1:25-CR-00504 (JSR), (S.D.N.Y. Feb 10, 2026).
[2] Warner v. Gilbarco, Inc., No. 2:24-cv-12333, at 11 (E.D. Mich Feb. 10, 2026)
The content of this McV Alert has been prepared for information purposes only. It is not intended as, and does not constitute, either legal advice or solicitation of any prospective client. An attorney-client relationship with McConnell Valdés LLC cannot be formed by reading or responding to this McV Alert. Such a relationship may be formed only by express agreement with McConnell Valdés LLC.
