Is a Party’s AI Research Privileged in Litigation? It Depends.

Is a party’s AI research privileged in litigation? It depends.

In February 2026, two federal courts reached seemingly opposite conclusions about whether use of AI tools by parties to litigation is protected by the attorney-client or work product privilege. One court found that a pro se plaintiff’s use of AI was privileged and another court found that a criminal defendant’s use of AI was not. 

In Warner v. Gilbarco, the U.S. District Court for the Eastern District of Michigan found that a pro se plaintiff’s use of AI in connection with litigation was protected work product. The defense moved to compel plaintiff to produce documents and information concerning her use of third-party AI tools in connection with the lawsuit after plaintiff interposed a work product objection to those materials.

The defense argued plaintiff had waived privilege by inputting the materials into a public AI platform that she used as a “drafting interface”. The court noted that Fed. R. Civ. P. 26(b)(3)(A) protects materials prepared by a party or its representative. Because a pro se litigant is the party, the materials she prepares in anticipation of litigation qualify based on the plain language of the rule. The court also noted that the waiver has to be to an adversary or one that would likely get into an adversary’s hand – merely using ChatGPT did not qualify.

On the same day, the Southern District of New York reached a different conclusion in United States v. Heppner. In Heppner, the court found that neither attorney-client privilege nor work product protection applied to AI generated documents created by a criminal defendant represented by counsel.

While a target of an investigation, Heppner used Claude to outline defense strategy and what facts and law he might argue if indicted. He later sent those documents to counsel. The government sought the documents in discovery and Heppner claimed privilege.

The court sided with the government, holding that the AI platform is not an attorney and therefore, the communications lacked the “trusting human relationship” required for privilege to apply; the communications were not confidential because the AI platform’s privacy policy permits data collection, training of the large language model, and disclosure to third parties including the government; and Heppner did not communicate with AI for the purpose of obtaining legal advice because the platform specifically says it does not provide legal advice (interestingly, Claude states: “I’m not a lawyer and can’t provide formal legal advice or recommendations”).

The court held that Heppner did not prove that he intended to obtain legal advice from the AI platform, or that his communications with Claude were prepared by or at the behest of counsel or reflect counsel’s strategies. The court suggested a different result may have been reached if counsel had directed the client’s use of the AI platform.

So, what can lawyers learn from these seemingly inapposite cases? First, we learn that the law is far from settled on these issues. The Warner court treats AI as a tool, the use of which does not result in privilege waiver, while to the Heppner court, AI is a third party, disclosure to which does waive privilege. Also, Warner maintained the distinction between attorney-client privilege (easily waivable) and work product protection (waived only by disclosure to an adversary). Heppner collapsed those distinctions. These issues are significant – they will ultimately determine whether inputs, prompts and outputs are privileged or potentially discoverable and under what circumstances. Therefore, we need to pay close attention to the developing law on these distinctions. 

Regardless, we can be sure that as AI becomes more ubiquitous, more and more of our clients will do their own research. Lawyers should counsel our clients about the potential unintended consequences of using AI without lawyer direction and supervision. Clients should not input protected materials to AI or use AI notetakers without lawyer direction.  

If the client is going to use AI, the lawyer should carefully coordinate and oversee that use - documenting that coordination and oversight - and the product ideally should be an enterprise level or closed platform, so that confidentiality is preserved.  

---

About the Author

Sharon Stuart, President & Claims Counsel

Attorneys Insurance Mutual of the South

Sharon Stuart is President and Claims Counsel of Attorneys Insurance Mutual of the South, Inc., a bar-related attorney malpractice insurance company, which was formed out of the Alabama State Bar by attorneys for attorneys.

Sharon is a founding partner of Christian & Small LLP in Birmingham, where she practices complex insurance, business, and class action litigation.

She is a former President of the Alabama Defense Lawyers Association, past President of the Birmingham Bar Association, and is Chair of the Alabama Supreme Court Standing Committee on the Rules of Civil Procedure. Sharon serves on the Alabama State Bar’s Board of Bar Commissioners and the Board of Legal Specialization. Sharon is a Fellow of the Alabama Law Foundation, the Birmingham Bar Foundation, and the American Bar Foundation.

Sharon is the 2025 recipient of the Tony McLain Professionalism Award, which recognizes outstanding, long-term service in the advancement of professionalism by a living member of the Alabama State Bar.