AI Platforms and the Risk of Privilege Waiver: Critical Lessons from United States v. Heppner

A recent decision out of the United States District Court for the Southern District of New York, United States v. Heppner, No. 25 Cr. 503 (JSR) (S.D.N.Y. Feb. 17, 2026), has significant implications for clients who access publicly available artificial intelligence platforms to analyze, prepare, or discuss matters related to litigation or legal advice. In a case of first impression, the court ruled that a defendant’s written exchanges with the publicly accessible generative AI platform Claude are not protected by either the attorney-client privilege or the work product doctrine. The decision underscores the substantial risks clients face when inputting confidential, privileged, or litigation-sensitive information into consumer AI tools, particularly those that collect user data to train their models.  It also serves as a reminder to clients of the importance of adopting AI use policies and training that set clear boundaries on acceptable AI usage and restrict the input of confidential information into AI tools.

The Heppner court identified several independent grounds for denying attorney-client privilege protection to the defendant’s exchanges with Claude. First, communications with an AI platform, which is not an attorney, cannot satisfy the fundamental requirement that a privileged communication occur between a client and its attorney. The court emphasized that “in the absence of an attorney-client relationship, the discussion of legal issues between two non-attorneys is not protected by attorney-client privilege.” Critically for clients considering AI use, the court also found that communications with Claude were not confidential because Anthropic’s privacy policy expressly states that the company collects data on users’ “inputs” and the AI tool’s “outputs,” uses such data to “train” the platform, and reserves the right to disclose user data to “third parties,” including “governmental regulatory authorities.” The court observed that users of such platforms cannot have any “reasonable expectation of confidentiality” when they have consented to these terms, and that if the defendant had input privileged information into Claude, doing so would have constituted a waiver by sharing the information with a third party. Additionally, the court noted that the defendant did not communicate with Claude for the purpose of obtaining legal advice—Claude itself disclaims providing legal advice and recommends users “consult with a qualified attorney.” Further, although the defendant shared his Claude exchanges with his counsel, non-privileged communications do not become privileged merely by being later transferred to an attorney. The court also held that work product protection did not apply because the defendant used Claude on his own initiative and not at the direction of counsel, and the documents did not reflect counsel’s mental processes or legal strategy.

While Heppner was decided in the Second Circuit, its reasoning draws on widely accepted elements of the attorney-client privilege—that protected communications must be between a client and attorney, be intended to be confidential, and be made for the purpose of obtaining legal advice—principles that are broadly consistent across federal courts and most state jurisdictions. It is also consistent with other recent judicial decisions finding limited privacy protections for users’ exchanges with generative AI, such as the January 2026 decision in In re: OpenAI, Inc. Copyright Infringement Litigation, 25-md-3143 (SHS) (OTW) (S.D.N.Y. Jan. 5, 2026), ordering OpenAI to produce millions of logs showing users’ exchanges with ChatGPT in a copyright infringement suit.

The Heppner decision carries several important lessons for businesses that use or are considering using AI platforms.

• First, clients should not input privileged or confidential information into consumer AI platforms, including information received from counsel, litigation strategy, legal theories, or any facts that form the basis of legal advice. Sharing privileged information with an AI platform could constitute a waiver, “just as if [the user] had shared it with any other third party.”
• Second, clients should carefully review AI platform terms of service and privacy policies; platforms that collect, retain, or use inputs and outputs for training purposes—or that reserve the right to disclose data to third parties—create significant confidentiality risks.
• Third, if AI tools are necessary, clients should use only enterprise or private deployments with contractual confidentiality protections, and such arrangements should be documented and reviewed by counsel.
• Fourth, any use of AI in connection with litigation or legal matters should occur only at the express direction of counsel.

The court’s decision confirms that using publicly available AI platforms for matters touching on legal advice or litigation exposes clients to substantial risk of privilege waiver. Whether other courts will follow the Heppner framework remains to be seen, but the decision serves as a clear warning to businesses: exercise caution before entrusting sensitive legal information to consumer AI tools. Businesses should proactively audit their use of AI platforms and consult with counsel to ensure that adopting these technologies does not inadvertently compromise legal protections.