When your expert witness admits to using ChatGPT in preparing their report, you're likely to face a discovery request for the entire conversation log. The Villanueva v. Las Vegas Metropolitan Police Department ruling provides a framework to respond, but only if you've documented AI use properly from the start.
Judge Anne R. Traum's decision clarifies which AI-generated content falls under Rule 26 disclosure requirements and which remains protected as work product. Your legal operations team needs a protocol now, before the next expert engagement.
Why This Matters Now
Expert Tom Melton used ChatGPT to research case law and draft portions of his report. When opposing counsel learned this during testimony, they subpoenaed his entire ChatGPT history log. The magistrate judge initially ordered production of all 40 pages, reasoning that Melton "considered" the AI output in forming his opinions.
Judge Traum reversed that blanket order. She found some portions discoverable under Rule 26(a)(2)(B)(ii) and others protected under work product doctrine. The distinction turns on whether the AI output constitutes "facts or data" the expert considered, or reveals the expert's mental impressions and litigation strategy.
If you're engaging experts who use AI tools, you're operating under this standard whether you've prepared for it or not.
What You Need Before Starting
Before your next expert engagement, assemble:
- Updated expert engagement letters requiring disclosure of all AI tool use, including specific platforms and dates.
- AI use logging template for experts to document each interaction: date, tool, purpose (research vs. drafting vs. deposition prep), and whether output was incorporated into deliverables.
- IT contact who can export chat histories from enterprise AI tools in a structured format.
- Counsel approval on your classification framework for AI interactions (see Step 2).
You'll also need buy-in from your expert witness coordinator and outside counsel managing expert discovery. They must understand that "we didn't know the expert used AI" is no longer a viable position.
Step-by-Step Implementation
Step 1: Mandate AI disclosure in expert agreements
Revise your standard expert engagement letter to include:
"Expert shall disclose all use of artificial intelligence tools, including but not limited to large language models, in researching, analyzing, or drafting any work product related to this engagement. Expert shall maintain a contemporaneous log of all AI interactions, including the tool used, date, purpose, and whether any AI-generated content was incorporated into deliverables."
Send this revision to all active experts immediately, not just new engagements.
Step 2: Create a four-category classification system
Train your experts and legal team to classify each AI interaction:
Category A: Case law summaries and legal research outputs
These are discoverable. Judge Traum found that AI-generated case summaries constitute "facts or data" under Rule 26(a)(2)(B)(ii). The expert considered them in forming opinions, even if they contained hallucinated citations. Log these interactions separately and plan to produce them.
Category B: Questions and prompts from the expert
These are protected. Judge Traum compared them to search queries, finding they reveal mental impressions rather than facts considered. Your expert's questions show their litigation strategy and areas of focus. Mark these as work product.
Category C: Deposition preparation materials
These are protected. When Melton used ChatGPT to create a deposition outline, Judge Traum found this revealed "focus, priorities, or areas of concern" for trial. It didn't supply facts supporting his opinion; it helped him prepare to defend it.
Category D: Draft report content
These are protected under Rule 26(b)(4)(B), which shields expert drafts "regardless of the form in which the draft is recorded." Judge Traum rejected the argument that AI-generated drafts fall outside this protection simply because the expert didn't personally type every word.
Step 3: Implement real-time logging
Don't rely on experts to reconstruct their AI use months later. Require:
- Screenshot or export of each substantive AI interaction within 24 hours.
- Brief description of purpose (which category above).
- Note of whether any output was copied into work product.
Store these logs in your matter management system, tagged to the specific expert and engagement. You'll need them when you respond to interrogatories about expert materials.
Step 4: Conduct privilege review before expert deposition
Two weeks before your expert's deposition, pull their AI use log and conduct a privilege review:
- Segregate Category A materials (plan to produce).
- Confirm Categories B, C, and D are properly marked as work product.
- Prepare a privilege log describing withheld AI interactions by category, not line-by-line.
- Brief your expert on how to answer questions about AI use without waiving protection.
Judge Traum noted that using ChatGPT doesn't waive privilege because "ChatGPT is not a person, but was an application used in this instance as a research, drafting, and editing program." You're not disclosing to a third party; you're using a tool. But you must still protect the work product that tool helped create.
Validation: How to Verify It Works
Test your protocol before you need it in litigation:
Scenario test: Have an associate use ChatGPT to research a hypothetical expert opinion issue. Can they correctly classify each interaction using your four categories? If they're uncertain, your framework needs clearer examples.
Log completeness check: Review three recent expert engagements. Can you identify from the file whether the expert used any AI tools? If not, your disclosure requirement isn't working.
Privilege log drill: Pull an AI use log and draft a privilege log entry for the protected portions. Does it describe the interaction's purpose without revealing the mental impression you're trying to protect? "AI-generated deposition preparation outline" works. "AI-generated outline addressing plaintiff's credibility concerns" reveals too much.
Maintenance and Ongoing Tasks
Quarterly: Review new AI tools your experts report using. Consumer ChatGPT, enterprise ChatGPT, Claude, Lexis+ AI, and Westlaw Precision each have different export capabilities. Update your logging template to accommodate new platforms.
After each expert deposition: Debrief on AI-related questions asked and how the expert responded. If opposing counsel is probing AI use more aggressively, tighten your deposition prep protocol.
Annually: Revise your expert engagement letter to reflect new case law. Villanueva provides the current framework, but courts will continue refining the boundaries as AI use becomes more common.
Judge Traum's ruling gives you workable categories, but only if you've documented AI use as it happens. The expert who can't remember which ChatGPT sessions informed their opinion versus prepared them for deposition has already lost the work product protection. Your protocol must make that distinction automatic.
