Expert witness caught using ChatGPT to clear 3M in $61M trial

If you are paying top dollar for an engineering expert witness to save your company from a multi-million-dollar lawsuit, the last thing you want is for that expert to ask ChatGPT to fabricate your innocence.

That is exactly what just went down in a high-stakes Houston courtroom. An engineering expert hired by industrial giant 3M used ChatGPT to draft almost his entire technical testimony, explicitly ordering the chatbot to clear his client of all blame.

The resulting digital paper trail turned a crucial defense into a courtroom catastrophe.

Prompting for 0% fault

The litigation stemmed from a deadly 2020 blast at the Watson Grinding manufacturing facility in Houston. The disaster killed three people and leveled roughly 200 neighborhood homes.

Federal investigators traced the explosion to a degraded, poorly crimped rubber welding hose that leaked flammable gas. Impacted homeowners sued 3M alongside Watson Grinding, alleging 3M failed to properly service the facility's gas detection equipment.

To deflect liability, 3M brought in Josh Autenrieth, an expert witness from Knighthawk Engineering.

Instead of conducting a purely independent evaluation, Autenrieth turned to OpenAI. Evidence revealed that ChatGPT wrote between 85% and 90% of his formal defense report.

His prompts were breathtakingly direct. He instructed the chatbot to craft an exceptional report defending 3M, explicitly demanding that the AI show how 3M is 0% at fault for the explosion.

350 pages of pure liability

Court discovery rules can be merciless, and Autenrieth quickly found out that private AI sessions do not stay private under subpoena.

The discovery process unsealed a staggering 350 pages of raw ChatGPT conversation logs between the expert and the bot.

Armed with the transcripts, plaintiff attorney Will Moye pulled off a devastating litigation move. He called 3M's own star witness to the stand as an adverse witness, systematically grilling Autenrieth on his prompts before the jury.

The strategy worked. The jury handed down a $61 million verdict, finding 3M 30% liable for the fatal disaster.

Why it matters

We have seen lawyers get sanctioned for citing hallucinated AI court cases, but this Houston explosion trial takes the intersection of generative AI and the legal system to a much darker place.

An expert witness is legally required to deliver independent, reliable technical analysis. When an expert openly prompts a language model with an outcome-driven mandate like prove my client has zero liability, that scientific credibility vanishes instantly.

It is also a massive wake-up call regarding legal discovery. Many professionals still treat chat boxes like private scrapbooks, forgetting that prompt logs are discoverable business records in civil litigation.

We think this case will force corporate legal teams to establish strict bans or disclosure protocols around generative AI for third-party consultants. Letting an expert wing a defense report through ChatGPT did not save 3M money; it handed the opposing counsel a 350-page roadmap to a $61 million payout.