{"id":906111,"date":"2026-09-10T16:27:53","date_gmt":"2026-09-10T22:27:53","guid":{"rendered":"https:\/\/www.myconstructionexpert.com\/blog\/?p=906111"},"modified":"2026-09-10T16:27:57","modified_gmt":"2026-09-10T22:27:57","slug":"expert-witness-compromised-by-artificial-intelligence","status":"publish","type":"post","link":"https:\/\/www.myconstructionexpert.com\/blog\/expert-witness-compromised-by-artificial-intelligence\/","title":{"rendered":"When AI Compromises Expert Witness Territory"},"content":{"rendered":"\n

Esquire Deposition Solutions<\/a><\/p>\n\n\n\n

Expert\u2019s use of artificial intelligence is continuing to create headaches for litigators. Recently a federal court in Connecticut concluded that\u00a0an expert witness\u2019s prompts into an AI tool were discoverable<\/a>, a ruling that seems likely to be widely embraced. And then there was the case of the\u00a0artificial intelligence expert whose report was contaminated by three non-existent, AI-generated citations<\/a>. The ironic result: he wasn\u2019t allowed to testify on AI\u2019s capacity to mislead.<\/p>\n\n\n\n

\n

Now that AI use is pervasive, and on every litigator\u2019s radar screen, cross-examination regarding AI should be expected.<\/strong><\/p>\n<\/blockquote>\n\n\n\n

Now come two more cases of expert misuse of generative artificial intelligence. The first, Laake v. 3M Company<\/em>, No. 2020-05505 (11th Dist., Harris County, Texas), is a negligence action arising from a gas explosion that killed three people and damaged hundreds of homes in 2020. One of the defendants, the 3M Company, retained an expert who (unbeknownst to 3M) relied on ChatGPT extensively while creating his report; one prompt asked ChatGPT to demonstrate that 3M bore zero percent of the fault for the gas explosion. The expert\u2019s use of ChatGPT use surfaced during his deposition, and his prompt history later became powerful impeachment material at trial. The jury ultimately allocated 30 percent of the fault to 3M and returned a $61.5 million verdict against 3M and another defendant.<\/p>\n\n\n\n

In the second case, LeDoux v. Outliers Inc.<\/em>, No. 3:24-cv-05808-TMC (W.D. Wash., July 24, 2026), the plaintiff alleged that nootropic supplements manufactured by the defendant caused her to fail a military drug test, triggering a court-martial and destroying her career. Defense counsel combed through the plaintiff\u2019s pleadings \u2014 the proposed amended complaint, a Daubert<\/em> motion, a motion to strike affirmative defenses, expert reports, and a response brief \u2014 searching for fabrications. Among the fabrications they uncovered were several errors in two expert reports. One report cited articles that either did not exist or carried the wrong authors, and a second expert\u2019s report repeated the identical errors. Unlike Laake<\/em>, however, LeDoux<\/em> did not involve experts independently turning to AI. Plaintiff\u2019s counsel admitted that she used Claude and ChatGPT to generate unverified material, including a citation table that she provided to two experts. Both experts incorporated the same erroneous citations into their reports without identifying the mistakes.<\/p>\n\n\n\n

Keeping a Closer Eye on Liability Experts<\/strong><\/p>\n\n\n\n

The credibility of the expert witness is often a key determinant of outcomes in civil litigation. It will be difficult to lean on the views of an expert witness if, in fact, it comes to light that the expert leaned heavily on AI when formulating his or her views. Fortunately, the Laake<\/em> and LeDoux<\/em> cases suggest a number of constructive measures that litigators might consider adopting in order to not get blindsided by embarrassing AI miscues. Among them:<\/p>\n\n\n\n