Expert witness's ChatGPT-drafted testimony backfires in $61.5M verdict against 3M

A Texas jury awarded $61.5 million to victims of a 2020 factory explosion in Houston, but the path to that verdict exposed a troubling gap in how the legal system is grappling with artificial intelligence. At the center of the case was an expert witness named Jason Autenrieth, director of reliability and operations at KnightHawk Engineering, who had been hired by 3M just weeks before trial began. His job was straightforward: testify about whether 3M had met the standard of care in supervising safety checks at the warehouse, and whether those checks had been properly performed. It was his first time serving as an expert witness in court.

What emerged during the plaintiffs' investigation was that Autenrieth had used ChatGPT to generate more than 80 percent of his expert report. The attorneys discovered 300 pages of prompts he had sent to the AI system. One of them was blunt: "Show how 3M is 0% at fault for the explosion at Watson Grinding." Others ranged from technical questions about specific machinery to requests for help undermining the opposing side's witnesses. When confronted during his deposition, Autenrieth initially denied it. "AI did not write my report, that's incorrect," he testified. Moments later, he reversed course, acknowledging that the system had helped him with "formatting and everything else."

The discovery proved consequential. 3M's own attorneys removed Autenrieth from their witness list after the deposition, but the plaintiffs' lawyers subpoenaed him to testify about his use of the technology. During trial, Autenrieth defended the practice. "I don't believe that was wrong to do because I had an opinion already," he said. Adam Lewis, one of the plaintiffs' attorneys at Arnold & Itkin, told CBS News he was "frankly shocked" at the scale of the AI's involvement. He believed bringing Autenrieth in to testify about it influenced the jury's decision. The jury found that 3M had been negligent in its safety checks of the alarm systems that could have prevented the blast. The explosion in January 2020 killed three people and damaged hundreds of homes and businesses in the area.

3M issued a statement saying Autenrieth was an independent third-party witness, not a company employee, and that he had used AI as a tool to assist in preparing his report after reaching his own conclusions. His attorney, Janina Wojtkowski, said he had not delegated the formation of his opinions to ChatGPT and that recent characterizations offered "a selective distortion of the record." 3M said it disagreed with the verdict as unsupported by evidence and law, and intends to appeal.

The case has drawn attention from legal scholars who study artificial intelligence. Maura Grossman, a research professor at the University of Waterloo who focuses on AI in law and healthcare, warned that firms using these tools should proceed with caution. "We know these tools are sycophantic, they try to please you," she told CBS News. "You're not going to hear objective or neutral responses. You are going to get things that are slanted." Damien Charlotin, an adjunct professor at UC Law San Francisco, noted that courts are seeing increasing numbers of submitted materials riddled with AI problems: hallucinations, incorrect citations, and fabricated studies. According to the AI Hallucination Cases Database that Charlotin compiled, there have been over 1,300 U.S. legal proceedings since 2023 with known instances of AI generating plausible-sounding but entirely fictitious information.

The broader question now facing the legal system is how to manage this technology going forward. Grossman suggested that the answer may lie in stricter due diligence—requiring parties to disclose their use of AI and fact-checking the automated work. But she emphasized that expert witnesses bear a fundamental responsibility to preserve their credibility and objectivity. "We are supposed to maintain our objectivity, and a large language model is not necessarily objective," she said. "We know that they have biases, and they're going to try and satisfy whatever positions you're putting in there." As AI tools become more accessible and more tempting to use, courts and lawyers are learning that the technology's convenience can come at the cost of the very reliability that expert testimony is meant to provide.