United States v. Heppner and AI Discovery: Confidentiality and Privilege Concerns

PDF

Recent trends across the country now demonstrate that a client’s use of public AI tools may be both discoverable and may potentially waive the attorney-client privilege.

On February 10, 2026, the U.S. District Court for the Southern District of New York reaffirmed foundational confidentiality principles in the context of artificial intelligence (“AI”) in the legal field in United States v. Heppner, No. 1:25-cr-00503-JSR.

In Heppner, the defendant was indicted on charges of securities and wire fraud in violation of 15 U.S.C. §78j(b) and 18 U.S.C. §1343, respectively, and related conspiracy, obstruction, and false statement charges. During the execution of a search warrant on his property, federal agents seized electronic devices containing AI logs and AI-generated documents, which contained the defendant’s use of AI to prepare reports outlining his legal defense and strategy.

The defense argued that the AI-generated documents were privileged on the basis that the AI-generated analysis was for the purpose of obtaining legal advice and was a means of consolidating the defendant’s thoughts to communicate them to counsel. Importantly, the defendant prepared these AI documents prior to retaining counsel and, therefore, not at counsel’s direction. In response, the government argued that the AI-generated documents were not confidential, necessarily disclosed to a third party, not created for the purpose of obtaining legal advice, and the defendant could not retroactively cloak the documents with privilege by later sending them to counsel.

The Court agreed with the government and ruled from the bench that documents generated using a consumer AI tool (even if later shared with counsel) were not protected by the attorney-client privilege or work-product doctrine because they were created outside counsel’s direction and via a third-party platform that disclaimed confidentiality. The Court’s analysis reflects a traditional waiver analysis: disclosure of confidential information to a third party generally destroys privilege, including disclosure to a public, open-source AI platform.

Just weeks earlier, another decision in the Southern District of New York upheld a discovery order requiring OpenAI to produce a sample of 20 million de-identified ChatGPT logs in pending copyright litigation, highlighting how courts are treating AI interaction data as discoverable evidence even where privacy protections are asserted. The District Court affirmed the magistrate judge’s orders requiring OpenAI to produce a 20-million-record sample of de-identified ChatGPT logs, including user prompts and model outputs, despite OpenAI’s efforts to narrow production or filter results. In doing so, the court held that OpenAI retained “tens of billions” of chat logs in the ordinary course of business and that these logs were discoverable. Taken together, these cases highlight that what clients (and perhaps even lawyers) type into generative AI platforms can become discoverable.

Beyond the federal context, state courts are beginning to weigh in on how AI use intersects with traditional duties of attorneys and litigants. Last year, the Supreme Court of South Carolina issued an Interim Policy on the Use of Generative Artificial Intelligence to provide guidance on responsible AI use within the judicial branch. While the policy formally applies to judges, court staff, and judicial employees, it also includes an explicit reminder that attorneys and litigants must ensure that AI does not compromise client confidentiality or otherwise violate the South Carolina Rules of Professional Conduct. Although the interim policy does not impose new obligations per se on private practitioners or litigants, it underscores a key point: courts are attentive to the ethical dimensions of AI use and to the accuracy and confidentiality associated with its use.

Especially in criminal cases such as Heppner, it is vitally important to the defense of the case that attorney-client communications remain confidential and privileged so that a charged client may have full and frank discussions with their counsel. The use of AI platforms to summarize information, provide advice, or even take notes during a meeting could be viewed as a waiver of privilege, and such a waiver could apply more broadly to entire subject matters of communications with counsel. Therefore, clients and attorneys must be proactive in thinking through the ramifications of the use of AI, especially in criminal cases. These developments signal that AI governance is now a matter of legal strategy and risk management, not just operational efficiency.