The Lawletter Vol. 51 No. 2
Becky Schanz—Senior Attorney
As recently noted by Judge Jed Rakoff, “[g]enerative artificial intelligence presents a new frontier in the ongoing dialogue between technology and the law.” United States v. Heppner, 25 Cr. 503 (JSR), 2026 U.S. Dist. LEXIS 32697, at *14 (S.D.N.Y. Feb. 17, 2026). Courts are just beginning to wrestle with the use of artificial intelligence (“AI”), but a few recent cases address whether privilege applies to AI use.
In Heppner, the district court held that the defendant’s conversations with an AI platform were not protected as work-product or by attorney-client privilege. Id. at *1. The defendant had been indicted on charges of securities and wire fraud and a search of his home produced his communications with a generative AI platform. Id. at *2-4. The defendant asserted that the communications occurred in preparation for a possible indictment and were privileged because they included information from his attorney, were to prepare for speaking with his attorney, and were ultimately shared with his attorney. Id. at *4-5.
Stating that attorney-client privilege protects "communications (1) between a client and his or her attorney (2) that are intended to be, and in fact were, kept confidential (3) for the purpose of obtaining or providing legal advice,” the court determined that the AI documents failed to meet at least two of the elements for the privilege to apply. Id. at *5-6 (quoting United States v. Mejia, 655 F.3d 126, 132 (2d Cir. 2011)). The generative AI platform was not an attorney and the communications were not confidential, both because the AI platform collected data on its users that it could disclose to third parties and because the AI platform itself is a third party. Id. at *6-7. The court further determined that the third element is also lacking because the defendant did not intend to obtain legal advice and was not directed by his attorney to use the platform. Id. at *8-9. The court noted that when the “Government asked [the AI platform] whether it could give legal advice, it responded that ‘I'm not a lawyer and can't provide formal legal advice or recommendations,’” before telling the user to consult an attorney. Id. at *9. The court also determined that the work product doctrine did not apply because the AI documents were not prepared by or at the request of an attorney nor did they show the defense counsel’s legal strategy. Id. at *11.
Around the same time, a magistrate judge in Michigan, in an employment discrimination case, denied the defendant’s discovery request for all information on the pro se plaintiff’s use of AI for the case. Warner v. Gilbarco, Inc., No. 2:24-cv-12333, 2026 U.S. Dist. LEXIS 27355, at *11 (E.D. Mich. Feb. 10, 2026). The court determined that the information was prepared in anticipation of litigation and was not discoverable. Id. The court rejected the defendant’s argument that the plaintiff waived her work-product privilege by using AI because the waiver had to be to an adversary and an AI program is not a person, but a tool. Id. at *12. The defendants also had no evidence that the plaintiff had uploaded confidential documents to the AI platform. Id. at *11 n.3.
In another employment discrimination case with a pro se plaintiff, the court held that the pro se plaintiff could protect his AI use, but had to disclose the platform he used to allow the defendant to determine if it provided sufficient protections for confidential information. Morgan v. V2X, Inc., Civil Action No. 25-cv-01991-SKC-MDB, 2026 U.S. Dist. LEXIS 67939, at *1-2 (D. Colo. Mar. 30, 2026). The court noted that both parties were using AI, but they disputed how it should be used with regard to confidential information. Id. at *4. The court distinguished Heppner because it was a criminal case and because the defendant acted on his own, without the advice of his counsel. Id. at *10. The court also reviewed Warner, finding it applicable because the plaintiff could also argue that his AI use was protected by the work-product doctrine. Id. at *10-11. The court found that simply using an electronic third-party system did not automatically waive the expectation of privacy. Id. at *12. Because AI platforms engage with the user, they invite the disclosure of information in a way that a general search engine did not. Id. The court held that the pro se plaintiff’s AI use was entitled to some work-product protection, but he must disclose which AI platform he used. Id. at *14-16.
Taken together, these cases suggest that it is unwise practice to discuss one’s legal problems with AI. The law on this issue is not fully developed, but there is a respectable chance that statements made to AI might be subject to discovery and admissible in court in at least some situations.
