Court Rules No Privilege Applies to Written Exchanges with Generative Artificial Intelligence Platform
April 27, 2026
By: Stephen B. Stern
In United States v. Heppner, Case No. 1:25-cr-00503-JSR (S.D.N.Y. Feb. 17, 2026), the United States District Court for the Southern District of New York ruled in a case of first impression that neither the attorney-client privilege nor the work product doctrine applied to protect written exchanges between an individual and a generative artificial intelligence (“AI”) platform.
In Heppner, Bradley Heppner was indicted and charged with a variety of crimes allegedly for defrauding investors in GWG Holdings, Inc. When FBI agents arrested him at his home in November 2025, they seized numerous documents and electronic devices pursuant to a search warrant. Among the materials seized were approximately thirty-one documents that memorialized communications between Heppner and a generative AI platform known as Claude (operated by a private company, Anthropic). Heppner argued that the AI related communications were protected by the attorney-client privilege and work product doctrine. He claimed the communications with Claude were made in 2025 “after [ ] Heppner had received a grand jury subpoena [and] after it was clear with discussions with the government that [ ] Heppner was the target of [an] investigation.” According to Heppner, the written communications reflected “reports that outlined defense strategy, that outlined what he might argue with respect to the facts and law that we anticipated that the government might be charging[,]” but these reports were prepared “[w]ithout any suggestion from counsel that he do so.”
With respect to the attorney-client privilege, the court found that the AI related communications lacked at least two, if not all three, elements that are required for the attorney-client privilege to apply. First, the court found that the communications were not between Heppner and his counsel, and the privilege requires communication between an attorney and client. Second, the court found that the communications were not confidential. The court reached this conclusion not solely because Heppner communicated with a third-party AI platform, but also because Claude’s written privacy policy, to which all users must consent before using Claude, disclosed that Claude collects data on both users’ inputs and outputs, it uses the data to “train” Claude, and Anthropic reserves the right to disclose such data to third-parties, including “governmental regulatory authorities.” Thus, the court concluded that Heppner could not have had a reasonable expectation that his communications with Claude would be confidential, and such an expectation of confidentiality is an essential element of a privileged communication. Third, the court found that Heppner did not communicate with Claude for the purpose of obtaining legal advice. Although the court acknowledged that its analysis surrounding the third element was a closer call because Heppner allegedly communicated with Claude for the “purpose of talking with counsel,” the court ultimately determined that Heppner’s intent to use Claude to communicate with counsel was not enough because Heppner’s communications with Claude were not made “at the suggestion or direction of counsel.” “Had counsel directed Heppner to use Claude, Claude might arguably be said to have functioned in a manner akin to a highly trained professional who may act as a lawyer’s agent within the protection of the attorney-client privilege.” Even then, the court was not clear that such direction would be sufficient, as it noted when Claude was asked if it could give legal advice, it responded by stating, “I’m not a lawyer and can’t provide formal legal advice or recommendations” and a user “should consult with a qualified attorney who can properly assess your specific circumstances.”
As for the work product doctrine, the court found that, even assuming the AI related communications were prepared in anticipation of litigation, which is an essential element of the work product doctrine, the communications were not “prepared by or at the behest of counsel” (they were prepared by Heppner on his own volition) and they “did [not] reflect defense counsel’s strategy.” Thus, the work product doctrine did not apply.
The court’s decision in Heppner is significant, as it is the first court decision regarding the protections, or lack thereof, concerning AI generated communications and analyses. While the decision in Heppner was made in the context of a criminal case, it likely will have ramifications in the commercial/civil context as well, and those ramifications may be far reaching. For example, individuals, whether in their personal capacities or as employees of a company, may conduct preliminary AI searches and analyses before discussing a particular matter with legal counsel, and those communications and analyses may need to be disclosed to an opposing party during discovery (whether through an forensic examination of an electronic device or otherwise), potentially giving the opposition insight into a variety of matters relevant to the litigation. To this end, attorneys should give consideration to including certain disclosures in litigation hold memoranda regarding the potential ramifications of AI searches. In addition, although some of the considerations analyzed in Heppner that weighed against finding the AI communications were privileged may not apply to an in-house attorney using an AI platform, there is no decision yet involving in-house counsel. Thus, it remains unclear whether in-house counsel’s use of a public AI platform (as opposed to a “closed” legal subscription platform such as Lexis, Westlaw, or Bloomberg) will be protected by the attorney-client privilege or the work product doctrine. This remains a new and developing area of the law that likely will continue to evolve. In the meantime, attorneys and clients should continue to monitor developments and clients (whether businesses or individuals) should communicate clearly with one another about potential uses and risks associated with AI generated communications and analyses.