How a ChatGPT user ‘embarrassed’ 3M in court

A Houston-based expert witness for 3M used ChatGPT to draft a key portion of his expert report in a lawsuit related to a 2020 explosion that killed three people and damaged more than 200 homes, according to court documents and legal filings. The revelation came during a deposition process when the prompts used by the witness, Josh Autenrieth of Knighthawk Engineering, were disclosed. These prompts showed that Autenrieth sought assistance from the AI tool to create an expert report that would defend 3M’s standard of care and assert that the company was 0% at fault for the incident. The case centers on the January 2020 explosion at Watson Grinding, a manufacturing facility in Houston. According to the U.S. Chemical Safety and Hazard Investigation Board, the explosion was caused by a degraded, poorly crimped rubber welding hose that allowed flammable gas to leak. Dozens of homeowners filed lawsuits against both 3M and Watson Grinding, alleging that 3M failed to adequately maintain the facility’s gas detection system and committed other errors contributing to the disaster. As part of the litigation, 3M engaged Autenrieth to prepare an expert report examining the company’s role in the incident. During the discovery phase, plaintiffs’ attorney Will Moye discovered a five-page document labeled “Citation Overlay,” which appeared to have been generated using artificial intelligence. Moye requested the prompts that Autenrieth had used to generate the report. This request led to a three-hour pause in the deposition as 350 pages of ChatGPT conversations were compiled. Among these materials were detailed instructions from Autenrieth, who informed the chatbot that he was representing 3M and required assistance in drafting an expert report supporting the company’s defense. Autenrieth uploaded numerous court records into ChatGPT and instructed the AI to analyze them and provide content that would bolster 3M’s position. The initial output from the AI was a roughly 30-page report stating, “From a technical and standard-of-care standpoint, 3M is 0% responsible for the January 24, 2020 explosion.” However, this statement was later removed after Autenrieth asked the AI to evaluate the report from the perspective of opposing counsel. Throughout the process, Autenrieth frequently revised the report using ChatGPT, asking the AI to refine its content and anticipate potential challenges from the opposing side. In one instance, ChatGPT awarded the report a score of 97 out of 100. Autenrieth also requested the AI to identify the top five issues that could be challenged and suggest counterarguments. Additionally, he inquired about specific components of a gas detector and questioned whether his qualifications as an expert witness were adequate given his limited prior experience in such roles. Despite acknowledging the use of ChatGPT, Autenrieth asserted that his professional expertise formed the foundation of his opinions. He stated, “I’ve got a body of work and 20-plus years of experience in the industry,” and explained that he used the AI to develop a preliminary framework, which he then adjusted based on his own knowledge. He further mentioned that he occasionally used the AI while commuting to jog his memory and assist with complex tasks. The final report submitted to the court largely mirrored the structure and content of the AI-generated initial version, with many sections directly copied from the original output. While Autenrieth admitted to relying on ChatGPT, he emphasized that his personal judgment and experience guided the final product. The case continues as the court evaluates the implications of AI-assisted expert testimony in legal proceedings.