AI-Generated Report Claims 3M Zero Responsible for Fatal Houston Explosion – Court Documents Reveal

A recent court case has exposed a shocking revelation: an expert witness for industrial giant 3M relied exclusively on artificial intelligence to craft a report declaring the company legally blameless in a deadly 2020 Houston explosion. The discovery, made during a deposition, reveals that Josh Autenrieth of KnightHawk Engineering—hired by 3M to defend the company’s standard of care—used publicly accessible ChatGPT prompts to generate a document stating 3M was “0% responsible” for the January 24, 2020, Watson Grinding explosion.

The incident occurred at a Houston industrial facility where a degraded and poorly crimped rubber welding hose leaked flammable propylene gas overnight, killing three people and injuring dozens. Federal authorities identified faulty equipment as the cause of the disaster. Following the explosion, 3M was contracted to inspect and maintain gas detection systems that allegedly failed, leading to multiple lawsuits from victims seeking compensation.

During a deposition, plaintiff’s counsel discovered Autenrieth’s expert report was built entirely on ChatGPT prompts uploaded online, with no human oversight. The lawyer confronted Autenrieth: “Well, do you realize, sir, that your chat right now is still publicly available?” Autenrieth admitted ignorance of how the AI functioned. Court documents later confirmed Autenrieth billed KnightHawk Engineering for 174 hours at $475 per hour—totaling at least $82,650—while his prompts explicitly instructed the AI to “show how 3M is 0% at fault” and conclude that the company “met the applicable standard of care.”

KnightHawk Engineering’s vice president defended Autenrieth’s actions as using AI merely as a tool, labeling recent claims of misconduct as “fake news.” Yet, the legal fallout has intensified: Harris County juries have already awarded $61.5 million to 24 victims in one case, with additional verdicts totaling over $150 million in damages to individual plaintiffs since June 2025. The revelation underscores growing scrutiny of how AI is increasingly leveraged in high-stakes legal contexts without adequate accountability.