3M Expert Asked ChatGPT for a Zero-Fault Defense. Jurors Assigned 3M 30% Blame.

A 3M-retained expert asked ChatGPT to help write a report showing the company was “0% at fault” for a deadly 2020 Houston explosion. A jury ultimately awarded plaintiffs more than $61 million and assigned 30% of responsibility to 3M. The gap between the requested conclusion and the verdict became visible because the expert’s ChatGPT conversations entered the case.

Josh Autenrieth of Knighthawk Engineering was hired by 3M to prepare an expert report in litigation over the January 24, 2020 Watson Grinding explosion. The blast killed three people. Autenrieth used ChatGPT to prepare significant portions of the report defending 3M, according to court transcripts, deposition documents and discovery material published by 404 Media.

The U.S. Chemical Safety and Hazard Investigation Board concluded that a hose disconnected inside a coating booth, releasing propylene that accumulated and ignited. The litigation includes claims that work on gas-detection equipment by Detcon, a former 3M subsidiary, contributed to the blast.

The conclusion was specified before the draft

The disclosed chats show Autenrieth told ChatGPT he had been retained to defend 3M, asked it to support the company’s standard of care, and instructed it to show that 3M was 0% at fault. He uploaded court records and asked the system to construct a defense that placed responsibility on Watson Grinding.

Plaintiffs’ attorney Will Moye found a document described as a five-page Citation Overlay that appeared to be generated by ChatGPT. After he sought the underlying prompts, discovery produced about 350 pages of ChatGPT conversations, including public links to full chats.

ChatGPT was used as drafter, reviewer and research aid

ChatGPT generated a roughly 30-page draft that said 3M bore no technical or standard-of-care responsibility for the explosion. That absolute language was not included in the report filed with the court after Autenrieth asked the model to review the draft from opposing counsel’s perspective.

The chats documented a wider role for the model

• Autenrieth repeatedly submitted revised report drafts to ChatGPT for edits and asked the model to grade the report and anticipate attacks on his qualifications and conclusions.
• He uploaded an image of a gas detector and asked ChatGPT to identify it and name industrial gas-detector models. Moye said the equipment was central to the dispute.
• Court records cited by 404 Media indicate that 3M paid Knighthawk roughly $90,000 for its analysis, while a 3M filing listed Autenrieth’s hourly rate as $475.

The expert and plaintiffs offered opposing accounts

Moye argued that Autenrieth had relied on ChatGPT not as an assistive tool but to form his opinions and write the report, and said the prompts were biased toward helping 3M win. Autenrieth testified that he supplied the opinions, used the model to create a draft to build from, and reviewed and altered output that did not reflect his views.

The verdict does not by itself establish how much the AI-assisted report affected the jury’s allocation of fault. But the case shows the litigation risk of putting generative AI into an expert’s workflow: the finished report may be only part of the record. The prompts, source materials, revisions and requests for strategic critique can become material for opposing counsel to examine.

One verdict inside a much larger fight

The broader Watson Grinding litigation remains active. In a quarterly filing for the period ending June 30, 3M said more than 1,900 claimants had brought allegations against multiple defendants. Earlier related trials reached different outcomes for 3M: juries found it partly liable in June and November 2025, then cleared it in an April 2026 trial.

Those different outcomes keep responsibility for the explosion contested across the wider litigation. In this trial, though, the disclosed prompts put Autenrieth’s method alongside his testimony that he independently supplied and reviewed the opinions in his report.