AI Use Backfires on Expert Witness in Texas Trial

Summary

A Texas trial over the 2020 Watson Grinding factory explosion has drawn attention to the risks of using artificial intelligence in expert testimony. Plaintiffs’ attorneys said Jason Autenrieth, an engineering expert retained by 3M, used ChatGPT to create more than 80% of his report and found about 300 pages of prompts. One prompt asked the system to show that 3M was “0% at fault,” while others sought technical explanations and help undermining an opposing witness. Autenrieth initially denied that AI wrote his report, then acknowledged that it helped with formatting, structure and drafting. He said he had already formed his opinions after reviewing thousands of pages and that ChatGPT did not replace his professional judgment; his lawyer called public descriptions of the record selective. 3M described him as an independent third-party witness, but its attorneys removed him from the witness list after the deposition. Plaintiffs subpoenaed him to explain the prompts, and their attorney said bringing him to testify affected the case. The jury later found 3M negligent in safety checks of alarm systems, and the victims received a $61.5 million award; 3M disputed the verdict and said it would appeal. Legal experts warned that large language models can produce sycophantic or slanted answers, hallucinations, incorrect citations and fake studies. One database has recorded more than 1,300 U.S. proceedings involving known AI hallucinations since 2023. Experts told CBS News that lawyers and courts may need stronger disclosure, fact-checking and due diligence, while expert witnesses remain responsible for preserving objectivity and credibility.