Client Alert: "It Depends: A Comparison of Recent Rulings Regarding AI Use and Privilege" by Katherine Charonko and Elizabeth Stryker

About the Authors:

Katherine E. Charonko is Bailey Glasser’s Electronically Stored Information & Technology Practice Group Leader who she provides strategic guidance in some of the nation’s most complex and consequential litigation and is part of national trial teams in state and federal courts across the country. Ms. Charonko is nationally ranked in the Chambers USA, USA Nationwide-Product Liability: Plaintiffs category and was named a 2026 Lawdragon 500 Leading Litigator in America. She was also named a 2026 winner of the Legaltech / ALM 2026 Monica Bay Women of Legal Tech Award. 

Elizabeth L. Stryker is a litigator and member of the firm's ESI & Technology Practice Group. Ms. Stryker was named a 2025 Lawdragon 500 X - The Next Generation as well as a Best Lawyers' Ones to Watch, among other recognitions.

Two federal courts have issued rulings regarding whether a party’s use of generative AI tools were protected by the attorney-client privilege or work product doctrine.

The answer, as attorneys are all too familiar with, is that it depends.

In February 2026, two courts reached different conclusions, with one (in United States v. Heppner, 25 Cr. 503 (JSR), 2026 WL 436479 (S.D.N.Y. Feb. 17, 2026)) finding that a defendant’s use of an AI tool was not privileged or protected, and another (in Warner v. Gilbarco, Case No. 2:24-cv-12333, 2026 WL 373043 (E.D. Mich. Feb. 10, 2026)) determining that a pro se litigant’s use of AI tools was protected by the attorney-client privilege and work product doctrine.

In the Second Circuit, a client’s use of generative AI tools to seek legal advice without direction from his counsel is likely not protected by the attorney-client privilege or work product doctrine. In Heppner, the court analyzed the attorney-client privilege in the context of the defendant’s use of AI tools to seek advice regarding a pending criminal investigation and found that the privilege did not apply for several reasons. The court opined that because the AI tool is not an attorney, then that alone disposed of his privilege claim. Heppner, 2026 WL 436479, at *2. Next, the Heppner court determined that the communications were not private and, therefore, not privileged. The court was persuaded by the AI platform’s privacy policy and found that the defendant had no reasonable expectation that his prompts to the AI tool would remain private. Id. at *3. Finally, the court examined the defendant’s actions and determined that he did not use the AI tool for the purpose of obtaining legal advice. Id.

The court also rejected defendant’s counsel’s assertion that the defendant had intended to share his findings with his counsel. The court noted that the attorney-client privilege does not later arise by sharing non-privileged communications with counsel. Id. Ultimately, the court ruled that the attorney-client privilege did not protect the communications between the defendant and the AI tool. The work product doctrine likewise did not protect the communications between the defendant and the AI tool because they were not prepared at the direction of the defendant’s counsel. Id. at *3–4.

In Warner v. Gilbarco, however, the court found that a pro se litigant’s use of AI tools was protected by the attorney-client privilege. The Warner court emphasized that the attorney-client privilege shields from discovery documents and tangible things prepared in anticipation of litigation or for trial by a party or its representative. Warner, 2026 WL 373043, at *11 (quoting Fed. R. Civ. P. 26(b)(3)(A)). The Warner court also determined that the information was subject to protection under the work-product doctrine, noting that waiver only applies when the communication is “to an adversary” or likely to end up with an adversary. Id. at *11–12 (citing In re Columbia/HC Healthcare Corp. Billing Pracs. Litig., 293 F.3d 289, 306 n.28 (6th Cir. 2002)). Per the court, an AI tool is not a person but an aid. Id. Finally, it is worth noting that the Warner court stated that the defendants’ “preoccupation with Plaintiff’s use of AI needs to abate.” Id. at *11 n.3. Courts may have little patience for the technical and unsettled nature of these kinds of discovery disputes, particularly where the fruits of these challenges may not be particularly relevant or important to the case.

Given these two recent rulings, the question of whether the attorney-client privilege and work product doctrine attach to a party’s use of AI tools remains unsettled. What is clear, however, is the need to continue to have additional and evolving conversations with clients regarding their use of AI tools. While a client may believe that they are helpfully providing their counsel with additional research or strategy, the client may be waiving any privilege or protection over these communications. See Heppner, 2026 WL 436479, at *3. To avoid this potentially harmful error, it is critical for counsel to ask their clients about their use of AI tools on a regular basis. As the use and development of AI expands, the legal questions surrounding privacy, privileges, and preservation, among other complex discovery issues will expand.

Bailey Glasser’s Electronically Stored Information & Technology Practice Group regularly encounters these types of issues. These attorneys counsel clients on the emerging pitfalls and advantages of the use of artificial intelligence in connection with legal proceedings. Representative work includes crafting policies and protective orders for clients, along with fulsome discovery support from written requests to the efficient use of databases and beyond.

Please contact the authors of this post if you need additional guidance on the use of generative AI tools in legal proceedings or regarding the assertion and maintenance of privileges and other protections.