In August 2026, a Harris County jury awarded $61.5 million to 24 plaintiffs in a lawsuit arising from the 2020 Watson Grinding explosion in Houston, which killed three people and destroyed roughly 200 homes. 3M, found 30% liable, had hired engineer Josh Autenrieth of KnightHawk Engineering as an expert witness.
Plaintiffs’ lawyers obtained 350 pages of his ChatGPT session logs through discovery. The logs revealed he had asked 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.”
How did the prompts become evidence?
The plaintiff’s lawyer took the unusual step of calling the opposing side’s expert as his own witness to demonstrate the AI misuse to the jury. ChatGPT session logs are now established as discoverable work product in US mass tort litigation. Every prompt, including instructions to bias the output, sat in a record the expert did not control.
This is the third plaintiff verdict across four Watson Grinding trials, with nearly 2,000 additional plaintiffs in the pipeline and the next trial set for October 2026. The case is not an isolated embarrassment. It is a repeatable discovery pattern.
What does this mean for expert witness AI governance?
An expert who uses a consumer chatbot to draft a report is creating a contemporaneous record of methodology. If that record shows advocacy rather than independent judgement, the record itself becomes the exhibit. Firms that cannot produce a governed audit trail will produce an ungoverned one instead.
Regulated professionals in legal and disputes work should treat every AI prompt as potentially producible. Check your governance exposure in three minutes.
This case is tracked in the Ecsper AI Risk Intelligence, which documents over 2,000 court cases involving AI-fabricated content.
