On January 24, 2020, a gas leak from a degraded and poorly crimped rubber welding hose ignited at the Watson Grinding and Manufacturing facility in northwest Houston, according to findings from the U.S. Chemical Safety and Hazard Investigation Board. The blast killed three people, including two Watson Grinding employees, injured 18 others, and destroyed or damaged roughly 200 nearby homes. Dozens of homeowners and victims sued both Watson Grinding and 3M, arguing that 3M — which had been contracted to service and inspect the facility's gas detection system — failed to properly maintain that system and missed warning signs for years leading up to the disaster. The litigation has since produced a string of bellwether trials, including a $118 million verdict in late 2025 and a $37.9 million verdict earlier in 2026, with nearly 2,000 additional plaintiff claims still pending in Harris County.
To defend itself in one of those trials, 3M retained Josh Autenrieth, an engineer with Knighthawk Engineering, to prepare an independent expert report on the standard of care surrounding the gas detection system. Autenrieth billed $475 an hour, and 3M ultimately paid Knighthawk roughly $90,000 for his 30-page report.
The unraveling of that report began with an easy-to-miss document. During discovery, plaintiffs' attorney Will Moye of the Moye Law Firm spotted a five-page attachment described as a "Citation Overlay," a document format Moye recognized as something ChatGPT generates automatically. Suspecting AI involvement, Moye demanded that 3M's legal team produce any and all AI prompts used in preparing Autenrieth's report. The request halted Autenrieth's deposition for three hours while the logs were gathered. What came back was staggering: roughly 350 pages of ChatGPT conversation logs, some of which even included public, shareable chat links that reporters at were later able to review directly.
Those logs showed Autenrieth essentially outsourcing his engineering judgment to a chatbot. He opened one conversation by telling ChatGPT, "I am being retained as a professional expert witness by 3M in defense of them in their lawsuits and other legal proceedings behind the January 2020 explosion at Watson Grinding," according to court transcripts cited by the New York Post. He then instructed the model to "create an exceptional expert witness report defending the standard of care at 3M" and to "show how 3M is 0% at fault for the explosion at Watson Grinding". Other prompts asked ChatGPT to "counter the defense witness['s] outlandish and false claims particularly about working on equipment you are not trained to and without the right permitting". Autenrieth fed the tool hundreds of pages of court records, his own CV, and, in a detail that drew particular ridicule from commentators, a photograph of a gas detector, the very device at the center of the case, along with the question "what am I looking at?".
ChatGPT produced roughly a 30-page draft report, and its first version bluntly concluded that "3M is 0 percent responsible for the January 24, 2020 explosion." When Autenrieth asked the chatbot to critique its own work by simulating opposing counsel, ChatGPT itself flagged the "0 percent responsible" language as "an easy target" for cross-examination, advice Autenrieth took, cutting the line before filing. He later asked the model to grade the finished report, and it scored itself 97 out of 100. Many of the prompts, according to Moye, were entered the night before Autenrieth's deposition.
At trial, Autenrieth conceded under questioning that somewhere between 85 and 90 percent of his report had been generated by ChatGPT. He defended his process from the stand, testifying that "my opinions were put in there, and AI helped me to draft a straw man to build off of," that he revised any output he disagreed with, and that his conclusions rested on "20-plus years of experience in the industry". Faced with the disclosure, 3M itself tried to disqualify its own expert. Moye responded by taking the unusual step of calling Autenrieth to the stand himself, walking the jury through the entire prompt-to-report pipeline.
Moye later told 404 Media that "this expert relied on AI not as an assistive device, but exclusively relied on ChatGPT to form his opinions and write his report," adding that Autenrieth "acknowledged [at trial] the prompts he put in were biased toward 3M to help 3M win the case... it's really egregious." Moye added a line that has since become the defining quote of the episode: "They hired him for the sole purpose of changing the outcome of the case. They hired him and he used ChatGPT to write these reports, so really, ChatGPT was the expert in the case. There's just no question about that". Knighthawk Engineering pushed back on the framing of the coverage as "selective and negative" and, in one public response, called it "Fake News," while maintaining that "Mr. Autenrieth utilized this technology as a tool to assist in reviewing questions, arguments, and counterarguments related to the case. AI did not replace his engineering judgment. Mr. Autenrieth developed and presented his own professional opinions and stands behind those opinions".
On August 10, 2026, a Harris County jury returned a $61,578,273.19 verdict for 24 plaintiffs, finding 3M 30 percent responsible for the explosion and assigning the remaining 70 percent to Watson Grinding, which had invoked its Fifth Amendment rights during the proceedings. It marked the third plaintiff-favorable bellwether verdict against 3M across four Watson Grinding trials, with the next trial set for October 2026 and roughly 2,000 additional claims still outstanding. 3M has said it disagrees with the verdict and intends to appeal.
The episode has quickly become a cautionary tale for anyone who manages expert witness engagements, and it raises questions that reach well beyond this one trial. Moye's ability to compel production of Autenrieth's 350 pages of ChatGPT history establishes, in practice, that prompts and drafts generated through AI tools can be treated the same way as an expert's notes, drafts, or communications: fair game in discovery. It also illustrates how an expert who instructs a chatbot to reach a specific, client-favorable conclusion before any real analysis has effectively pre-determined the "opinion," precisely the kind of bias that cross-examination is built to expose. Commentators have also pointed out that retaining firms and litigation support teams share in the exposure here, since it remains unclear whether 3M's engagement agreement with Autenrieth addressed AI use at all, and whether firms that vet or place experts should now be writing AI-disclosure terms into retention agreements up front rather than discovering the issue mid-deposition. Finally, as Plagiarism Today observed, an expert being challenged, or even disqualified, by their own retaining party over undisclosed AI use is often a "death knell" for that expert's future credibility in litigation, regardless of how any single trial turns out.
The case is likely to be cited for years as litigators, expert witness networks, and courts work out clearer rules for AI disclosure, vetting, and discovery in expert testimony.