3M Company

(New York Stock Exchange)
Back to all articles
Bearish -65

Expert Used “0% Fault” Defense ChatGPT Prompt, Defending 3M in Deadly Explosion Case

AI-generated analysis

📉 3M faces ongoing litigation over a 2020 explosion at Watson Grinding & Manufacturing that killed three people and caused roughly $200 million in property damage.

⚖️ The central legal allegation is that 3M failed to detect an 'open loop fault' in the gas detection system it was contracted to inspect and maintain.

🤖 Engineer Josh Autenrieth, retained by 3M, admitted using ChatGPT to draft approximately 85–90% of his expert witness report for the case.

📄 Plaintiffs' attorney Will Moye discovered AI-generated content in a 'Citation Overlay' document and demanded all prompt logs during a three-hour deposition pause.

💰 Juries have consistently found 3M partially liable, awarding damages totaling over $217 million across multiple bellwether trials.

📉 One recent jury assigned 3M 30% fault for the explosion, resulting in a $61.5 million verdict.

⚖️ Another trial found 3M 49% at fault, leading to an $118 million award against the company.

🇨🇦 A Canadian judge ruled that AI-generated expert evidence must carry no weight because it lacks a known individual with specific knowledge.

⚠️ Legal experts warn that prompt logs are now discoverable and may expose companies using AI for substantive analytical work in litigation.

📉 3M's defense strategy involving AI generated arguments claiming the company was '0% responsible,' which were later flagged as weak points by the AI itself.

Risk Factors
  • 3M is facing multiple jury verdicts assigning it partial liability for a deadly explosion, with damages awarded totaling over $217 million across several trials.
  • The company's gas detection system was alleged to have an 'open loop fault' that failed to detect a propylene gas leak leading to the January 2020 disaster.
  • Retained expert witness Josh Autenrieth used ChatGPT to draft approximately 85–90% of his expert report, raising concerns about the authenticity and independence of the defense analysis.
  • Plaintiffs' counsel successfully produced 350 pages of ChatGPT conversations and prompts, exposing the extent of AI usage in the company's litigation strategy.
  • A Canadian judge explicitly stated that AI-generated expert evidence must be given 'no weight,' setting a precedent that could impact future proceedings involving 3M.
  • The use of AI to draft legal arguments claiming the company was '0% responsible' was flagged by the AI itself as an 'easy target' for cross-examination, undermining the defense's credibility.
  • Bellwether trials have established a pattern of significant liability for 3M, with one jury finding the company 49% at fault and another assigning 30% responsibility.
Full Analysis
3M is facing significant liability challenges stemming from a deadly explosion at Watson Grinding & Manufacturing on January 24, 2020, which killed three people and caused approximately $200 million in property damage. The incident involved a propylene gas leak that ignited when a worker flipped a light switch, but the core legal dispute centers on whether 3M failed to detect an 'open loop fault' in the facility's gas detection system it was contracted to maintain. During litigation regarding this disaster, 3M retained engineer Josh Autenrieth to submit an expert witness report arguing the company was 0% at fault. However, discovery revealed that Autenrieth utilized ChatGPT to draft approximately 85–90% of his 350-page report, including uploading court records and photos for analysis. Plaintiffs' attorney Will Moye identified AI-generated artifacts in the documents, leading to a three-hour deposition pause where all prompts and conversations were produced. Despite the heavy reliance on AI for drafting defense arguments, juries have consistently found 3M partially responsible for the explosion. Verdicts from bellwether trials assigned 3M 30% fault with a $61.5 million award, 49% fault with an $118 million award, and another resulting in a $37.9 million judgment. While a separate 2026 trial cleared 3M entirely, it remains an outlier compared to the prevailing trend of liability findings. The case highlights growing scrutiny regarding the use of AI in high-stakes legal contexts, with Canadian judges already placing 'no weight' on AI-generated expert evidence. Legal observers warn that if AI performs substantive analytical work, discovery will uncover prompt logs, potentially unraveling expensive expert engagements and blurring the line between independent analysis and machine-generated advocacy.