Thought Leadership

For more than a century, civil law in the United States has protected the communications of litigants under a number of protections, including attorney client privilege and the work-product doctrine. That longstanding framework began to shift on February 17, 2026, when Judge Rakoff of the United States District Court for the Southern District of New York took the first step toward modernizing privacy and confidentiality rules to address the ever looming issue of generative AI.

In the landmark case of United States of America v. Bradley Heppner, 820 F. Supp. 3d 292 (2026), Judge Rakoff held that the documents a criminal defendant generated using Claude, a publicly available generative AI platform operated by Anthropic, were not privileged or protected as work product. The facts of Heppner include an indictment for a variety of alleged frauds and falsification of corporate records, which arose out of his alleged misconduct as an executive of multiple corporate entities. In conjunction with Mr. Heppner’s arrest, the FBI executed a search warrant of his home and seized numerous documents, including 31 documents which memorialized his communications with Claude. These documents covered particularly pertinent topics, such as defense strategies, possible arguments to specific facts or law he anticipated to be included in the indictment, and were prepared without any suggestion from his counsel to do so.

By and through his counsel, Mr. Heppner asserted privilege over the documents generated by Claude. He argued (1) that the documents included information he had learned from his attorney; (2) were created for the purpose of speaking with counsel to obtain legal advice; and (3) had all been shared with counsel after they were created. Following oral argument on the government’s motion seeking a determination that the AI-generated documents were neither protected nor privileged, the court granted the motion on February 10, 2026.

In issuing his opinion, Judge Rakoff determined that the AI documents lacked two, “if not all three,” of the elements required for attorney-client privilege—namely that the communications were not “between a client and his or her attorney,” the documents were not “intended to be” or were “in fact” kept confidential, and the AI documents were not prepared by Mr. Heppner “for the purpose of obtaining legal advice.” Put simply, Judge Rakoff noted that, “Because Claude is not an attorney, that alone disposes of Heppner’s claim of privilege.” Ultimately, Judge Rakoff concluded that, as the generated documents were not prepared at the request of Mr. Heppner’s counsel and did not disclose his attorney’s strategies, they failed to merit protections as work-product and were thus discoverable by the government.

The ramifications of Heppner will likely cause many changes in litigation practices as a direct result of the increased use of generative AI. Indeed, use of generative AI has exploded as of late. Recent data from Sciencedirect.com shows that the top 40 generative AI tools received nearly 3 billion monthly visits worldwide in March 2024 alone. Even Judge Rakoff noted the ever-growing prevalence of AI use, stating that “more than half of United States households have adopted AI in some form.”

While Heppner involved a criminal case, the holding will similarly be applicable to civil matters. It is incumbent on health care litigators to thoughtfully engage in discovery to determine whether plaintiffs consulted a generative AI platform in relation to their case. In practical use, this could be completed by questioning plaintiffs at deposition regarding whether they have asked ChatGPT, Claude, or one of the dozens of other generative AI platforms regarding the possible value of their cases, advice on what to say during a deposition, or a number of other queries. Likewise, these questions could be directed to a plaintiff via an Interrogatory or Request for Admissions. Regardless the avenue of questioning, it is paramount for attorneys to test the boundaries of courts across the country which are only now braving the new world intersections of AI, privacy, and privilege.