Are AI Communications Protected by Attorney–Client Privilege? Lessons from United States v. Heppner

A recent decision from the United States District Court for the Southern District of New York provides one of the first judicial examinations of whether communications with generative artificial intelligence (“AI”) platforms can fall within the protections offered by the attorney–client privilege or the work product doctrine. In United States v. Heppner, 25 Cr. 503 (JSR) (S.D.N.Y. Feb. 17, 2026), the court held that they do not.

This decision arrives at a critical moment where AI is being integrated into all aspects of our lives. Heppner makes clear that the efficiencies offered by AI come with legal risk.

Communications with AI platforms are discoverable, and it is therefore important to limit such communications to preserve the confidentiality of legal advice and strategy.

Although the case arose in a criminal context, the court’s reasoning applies equally to all aspects of legal representation.

The Court’s Decision in United States v. Heppner

In Heppner, the Government sought access to written exchanges between the defendant and a generative AI platform. The defendant had used the platform to ask legal-related questions. His attorneys then argued those communications were protected by attorney–client privilege and the work product doctrine. The Court rejected those arguments, finding that neither protection applied.

The Court grounded its decision in settled privilege law. Attorney–client privilege protects confidential communications between a client and an attorney made for the purpose of obtaining legal advice. The defendant’s communications did not meet those requirements. The Court emphasized that generative AI platforms are not attorneys and do not provide legal advice. Rather, they generate responses based on predictive algorithms trained on large datasets. While those responses may resemble legal analysis, they do not establish an attorney–client relationship or transform the platform into a legal advisor.

The Court also stressed that confidentiality is essential to privilege. First, the Court looked at confidentiality through the lens of whether there was a “reasonable expectation of privacy.” The Court said no such expectation existed.

To reach that conclusion, the Court reviewed the privacy policy of the AI platform at issue, “Claude” developed by Anthropic. The Court determined that no reasonable expectation of privacy existed because Anthropic collects user inputs and outputs to train Claude and reserves the right to disclose that data to third parties.

Second, the Court also rejected the application of the Kovel doctrine, which allows communications made to third parties to remain privileged where the third party is necessary to facilitate legal advice and is acting under the attorney’s direction. In doing so, a key fact was that defendant communicated with the AI entirely on his own. The Court did, however, leave open the possibility of Kovel doctrine’s application to AI communications, but only if the client is first directed to use the platform by their attorney.

The Court also rejected application of the work product doctrine, which protects materials prepared by or for an attorney in anticipation of litigation. Because the communications were not created by an attorney or at an attorney’s direction, and the AI platform was not assisting counsel in litigation preparation, the doctrine did not apply.

Importantly, the Court did not create new rules specific to artificial intelligence. Instead, it applied longstanding privilege principles. The decision confirms that emerging technologies do not alter the core requirements of privilege: an attorney–client relationship and the preservation of confidentiality.

Broader Implications and Key Takeaways

Although Heppner was a criminal case, its broader significance lies in its application of universally applicable privilege doctrines. Indeed, the Court applied principles that are equally applicable to all areas of the law. As a result, the decision’s impact extends across the full spectrum of legal practice.

The key takeaway is that this case highlights the need for attorneys and businesses to be proactive in risk management.

For businesses, this means carefully evaluating the AI platforms they use. Companies should not assume that AI tools provide confidentiality protections equivalent to communications with counsel. Instead, businesses should conduct thorough reviews of platform terms of service, privacy policies, and data handling practices to determine whether information entered into the platform may be stored, accessed, or disclosed. Without appropriate safeguards, sensitive legal information may become discoverable.

For attorneys, the decision underscores the importance of advising clients about the risks associated with AI use. Clients should be cautioned against inputting legal advice, attorney communications, draft pleadings, internal investigative findings, or other sensitive legal documents into AI platforms, particularly those operated by third parties. Doing so may undermine privilege protections and expose those materials to discovery.

Ultimately, Heppner reinforces a fundamental principle: privilege depends on confidentiality and the attorney–client relationship. As AI becomes an increasingly common tool in everyday life, preserving privilege will require deliberate and informed risk management.

Conclusion

The Southern District of New York’s decision in United States v. Heppner represents a significant development in the evolving relationship between artificial intelligence and legal practice. By holding that communications with generative AI platforms are not protected by attorney–client privilege or the work product doctrine, the Court has clarified an issue with immediate and widespread practical consequences.

Heppner is among the first cases applying traditional privilege doctrines to emerging technology, and more decisions are likely to follow.