Villanueva v. Las Vegas Metropolitan Police Department, No. 2:24-cv-00125-ART-DJA (D. Nev. Sept. 29, 2026), full opinion (PDF)
A federal district judge in Nevada has ruled that a testifying expert must disclose the ChatGPT case summaries he consulted in forming his opinion. The ruling addresses which parts of a testifying expert's AI history the opposing party can obtain. In a civil rights suit, the plaintiff's expert had used ChatGPT to research and write his report. A magistrate judge then ordered the expert's entire history log produced. The district court narrowed that order by protecting the expert's prompts, a deposition outline, and the material he used to draft his report.
What happened
The expert first revealed his use of ChatGPT at his deposition, although an earlier subpoena had sought all information he considered in forming his opinions. After the deposition, the plaintiff refused to produce the history log, claiming work product and trial-preparation protection. The plaintiff described the log as a mix of case research, deposition preparation, and drafting.
The magistrate judge ordered all of it produced. He relied on the plaintiff's description of the log without reviewing it in camera.
The court's analysis
Reviewing the log in camera, the district court accepted the magistrate judge's reliance on the plaintiff's good-faith description. Even so, it concluded that ordering the entire log produced was clearly erroneous or contrary to law. That deferential standard governs a district judge's review of a magistrate judge's discovery orders.
The court sorted the log by how the expert used each part. In its view, output an expert consults in forming his opinion is facts or data that Rule 26(a)(2)(B)(ii) requires the expert report to disclose. The court then protected the prompts, the deposition outline, and the drafting material, each with its own reasoning.
The court read Rule 26(a)(2)(B)(ii) to reach material an expert consulted as well as material he relied on. Because this expert requested the ChatGPT case summaries and consulted them in reaching his opinion on excessive force, the court concluded they had to be produced. It accepted the defendants' argument that the log was the only place to check for hallucinations in the citations and quotations the expert considered.
The court treated a prompt as a question the expert asks, much like an online research query. A question, in its view, sits outside the facts or data an expert considers. The court also accepted the plaintiff's argument that the prompts reveal the expert's mental impressions and theory of the case. It treated the prompts as protected under Rule 26(b)(3)(B).
According to the court, the expert used the deposition outline to prepare for his deposition and trial, not to inform his opinion. Rule 26(b)(3)(B) barred production, the court concluded, because the outline would reveal his focus, priorities, or areas of concern.
For the drafting material, the court applied Rule 26(b)(4)(B), which protects an expert's draft reports in whatever form they are recorded. Because the expert testified that he sometimes pasted ChatGPT's outputs into his report, the court treated those parts of the log as drafts. Even without that treatment, it added, they would be protected as his mental impressions.
The court also concluded that any privilege survived the expert's use of ChatGPT. It reasoned that ChatGPT is an application used for research, drafting, and editing rather than a person. Two other objections, neither involving ESI, were overruled.
Why it matters
Counsel retaining an expert should learn early whether and how the expert used generative AI. Villanueva treats AI-generated summaries an expert consulted as facts or data. Producing them with the expert report is the safer course. Early disclosure can narrow any later dispute to the prompts, preparation material, and drafting material that Villanueva protected.
An expert who uses these tools for research, deposition preparation, and drafting should keep records showing which outputs informed the opinion. When one log mixes those uses, those records can help a court decide where each entry falls.
A party seeking an opposing expert's AI history should ask for the outputs the expert consulted, the part of the log Villanueva treated as discoverable. Those outputs give the party the material it needs to test the expert's research for hallucinated citations. Challenges to the expert's use of the tool can then be developed for impeachment at trial.
Villanueva is one district court's ruling on a new question. Other courts may decide differently, particularly for outputs an expert used both to research and to draft.
The full opinion is available as a PDF.
