A New Frontier: AI and Privilege

The rapid proliferation of artificial intelligence has transformed our lives and provided a powerful tool that can enhance efficiency, streamline research, and expand access to information. At the same time, AI integration has created significant uncertainty, particularly in the legal field. Courts have increasingly been called upon to confront issues pertaining to the use of AI and have grappled with how traditional legal principles apply in this new frontier.

A recent decision from the U.S. District Court for the Southern District of New York provides a good example. On February 17, 2026, the court ruled that certain written exchanges that defendant Benjamin Heppner had with a generative AI platform were not shielded from discovery by the attorney-client privilege or the work product doctrine.

Background

On October 28, 2025, a grand jury returned an indictment charging Heppner with securities fraud, wire fraud, and conspiracy to commit securities fraud and wire fraud, among other charges. Heppner was arrested in November 2025, and agents from the Federal Bureau of Investigation executed a search warrant at Heppner’s home and seized numerous documents and electronic devices. During the search, the FBI seized documents that memorialized conversations Heppner had with the generative AI platform called “Claude.” Claude is operated by a private company called Anthropic.

Heppner’s counsel represented that the communications took place after Heppner had received a grand jury subpoena and after it was clear from discussions with the government that Heppner was the target of an investigation. Heppner used Claude to prepare reports that outlined defense strategy and what he may have argued based on the anticipated charges the government might have brought. Heppner insisted that the communications and reports were made in anticipation of a potential indictment. Heppner claimed that these documents were privileged.

SDNY’s Decision

The government moved for a ruling that the AI documents were not protected by the attorney-client privilege or the work product doctrine. The court agreed, holding that the documents were not protected by the attorney-client or work product privileges, and granted the government’s motion.

First, the communications were not between Heppner and his attorney, and Claude was not an attorney. Those facts alone defeated any claim of privilege.

Second, the communications with Claude were not confidential. The written privacy policy and consent to use Claude stated that Anthropic collected user data to train Claude and that Anthropic had the right to disclose such data to third parties, including the government. For that reason, Heppner had no reasonable expectation of confidentiality in those communications with Claude. The court did not believe that the AI documents shared any similarities to confidential notes that a client prepares to share with an attorney. Heppner shared the equivalent of his notes with Claude, a third party, before meeting his attorney first.

Third, although the court thought it was a close call, Heppner did not communicate with Claude for the purpose of obtaining legal advice. Claude disclaimed giving legal advice. In fact, when the government asked Claude if it could provide legal advice, it responded that it could not provide formal legal advice or recommendations and encouraged the user to consult a qualified attorney. The fact that Heppner shared the documents with counsel later did not change them into privileged documents — documents do not obtain privilege because they are transferred to legal counsel.

The documents also did not qualify for protection under the work product doctrine. The documents were not prepared by or at the direction of counsel, they did not reflect defense counsel’s strategy, and they were prepared on Heppner’s own volition. Heppner’s counsel admitted that the documents did not reflect strategy at the time they were created.

Practical Implications

This case serves as a stark warning about the risks of using AI tools. Documents generated by generative AI can be discoverable, even when the user is seeking legal information. Courts can find that users have no reasonable expectation of privacy in documents and information from AI tools, particularly where the platform’s terms allow data use or disclosure. Employers across all industries should strongly caution their employees about using AI and the lack of privacy or protection afforded to those inputs. Indiscriminate use of AI can increase liability exposure for individuals and employees alike and operate as a waiver of any potential legal privilege.