A Nevada federal court has issued one of the first rulings on when AI-generated research by an expert witness must be shared in litigation. In Villanueva v. Las Vegas Metropolitan Police Department, Judge Anne Traum considered whether the defense could compel plaintiff's expert Tom Melton to produce his entire ChatGPT history log. Melton had used the AI tool to research and draft his expert report on excessive force, including asking for case summaries of controlling precedents.
The court drew a careful line. Under Rule 26(a)(2)(B)(ii), experts must disclose "facts or data" they considered in forming their opinions—not just what they relied on. Citing Ninth Circuit precedent, Judge Traum found that the AI-generated case summaries were exactly that, because Melton consulted them as a quick-reference guide. Since the history log is the only place to review alleged hallucinated citations and quotes, those responses must be produced.
However, the court refused to order production of the entire log. The expert's questions and prompts to ChatGPT are not "facts or data," and are protected as opinion work product—either as part of a draft report or trial preparation materials. The same protection applied to Melton's use of ChatGPT to create a deposition outline. The ruling thus gives litigants a framework: AI outputs an expert actually considers may be discoverable, but the thought process behind querying the tool remains shielded.