What to Know
- ChatGPT use by expert witnesses is creating a new frontier for litigation discovery.
- AI chat logs may expose expert reasoning and become key evidence in court disputes.
- Expert engagement terms should address AI use, recordkeeping and independent verification.
In Montenegro Enterprises v. 3M Co., an ongoing multidistrict litigation involving liability for a Houston gas explosion, an expert witness used ChatGPT to formulate most of his expert opinion, highlighting new risks associated with AI use in litigation and expanding both offensive and defensive discovery considerations.
According to an August 17 report, an expert hired by 3M to defend the company in the cases — where the plaintiffs allege that 3M is responsible for causing the 2020 explosion that killed three people in Harris County, Texas — asked ChatGPT to “create an exceptional expert witness report defending the standard of care at 3M” and to “show how 3M is 0% at fault.”
Opposing counsel discovered the expert’s use of artificial intelligence during his deposition, resulting in the production of more than 300 pages of ChatGPT conversations. Ultimately, when he was cross-examined at trial, the expert conceded that up to 90% of his report had been written by AI.[1]
While this case is getting media attention because of the fatal explosion and the discovery that the expert used AI, it serves to highlight the fact that no bar association, court or professional body has set clear rules for how experts may use generative AI to formulate their opinions.
Though there could be a benefit to doing so, there is no doubt that experts are using these tools with no guardrails in place, nor any consistent standards for what must be disclosed, preserved or produced.
Many lawyers have already begun weaving requests for AI logs into fact and expert discovery, and those requests are on track to become a standard feature of discovery demands as the technology finds its way into nearly every facet of the legal profession. But all litigators who retain experts need to plan for this discovery now, and not after their expert is confronted with an AI chat log at a deposition or trial.
A New Discovery Frontier
Generative AI creates categories of records that simply did not exist in the mainstream a few years ago. Just like mobile devices made requesting content from applications like iMessage and WhatsApp nearly boilerplate, demands for generative AI input and output will no doubt become the norm.
Experts’ working processes have historically been invisible, protected in practice by the simple fact that these processes often were not recorded. But AI presents a sea change, because every prompt is a time-stamped record of what the expert asked for and, sometimes, the outcome the expert was aiming to achieve before the report was signed and submitted.
As trial lawyer Chris Schwegmann explained in a recent American Bar Association publication, these chat logs “are the closest thing civil discovery has ever seen to a transcript of the unguarded mind.”[2]
This means that when an expert uses AI, the process could be exposed through discovery, and counsel should be prepared to seek from every expert full session and prompt logs, system instructions given to the AI, and any AI-assisted test exercises or feedback on the expert’s conclusions.
Notably, courts are already beginning to endorse these kinds of requests. In a May decision in Conservation Law Foundation Inc. v. Shell Oil Co., US Magistrate Judge Thomas O. Farrish in the US District Court for the District of Connecticut ordered the “disclosure of any artificial prompts and/or queries used by [the expert] or her team in the course of producing her expert witness report.”[3]
In so holding, the court found that the parties’ Rule 29 of the Federal Rules of Civil Procedure discovery stipulation, providing that there would be no discovery of each other’s “expert notes, drafts, or communications,” was not specific enough to shield the AI materials.
Navigating the AI Minefield
Given these risks, counsel engaging experts need to make clear to them, early and often, that opposing counsel will request and may receive their AI conversations, and that those conversations could severely undermine the expert’s work at a deposition or trial.
To further protect themselves and their experts, counsel should consider modifying expert engagement letters to address AI, such as:
- A requirement that the expert tell retaining counsel about the AI tools that may be used;
- An approval step mandating the expert seek counsel’s permission before using AI;
- Recordkeeping provisions ensuring that prompts, iterative refinements and outputs are preserved and not overwritten; and
- A verification requirement that the expert independently validate any AI-generated content against the underlying facts and the expert’s — human — expertise.
If possible, parties to civil litigation should also consider stipulating, pursuant to Rule 29, that neither side’s expert’s AI usage will be discoverable. That said, given that AI use could completely undermine an expert’s work, parties may be hesitant to cut off this line of inquiry.
Having It Both Ways
Despite these risks, the universal embrace of AI is likely to mean that experts will find ways to integrate it into their practice. And perhaps there’s good reason to do so, as AI is undoubtedly a powerful tool that experts could use to play devil’s advocate, run mock cross-examinations or grade the report against the other side’s expert. Used well, these tools can genuinely strengthen an opinion.
Of course, an expert should still be able to show that despite the use of AI, the opinion still reflects the expert’s own independent judgment; that the expert verified what the AI produced against the facts; and that the expert can point to clear documentation of the reasoning that connects the evidence to the conclusion.
[1] https://www.404media.co/show-how-3m-is-0-at-fault-expert-witness-used-chatgpt-to-write-report-defending-company-in-deadly-explosion-lawsuit/.
[2] Schwegmann, Chris. “The Unguarded Mind: AI Chat Logs and a New Frontier for Discovery.” Litigation, vol. 52, no. 3, Spring 2026.
[3] Conservation Law Foundation, Inc. v. Shell Oil Co. et al., No. 21-cv-00933 (D. Conn) (Dkt. No. 970).