Lex Blog
Today’s guest post is from Dechert‘s Mary Kim. She takes a closer look at a case where plaintiff’s use of AI landed them in hot water and how defendants can make the most of these situations. As always our guest posters deserve 100% of the credit (and any blame) for their writings.
The existing commentary on the growing epidemic of artificial intelligence (“AI”) hallucinations in legal filings has largely focused on the forest—the staggering and accelerating volume of cases, along with the inadequacy of existing sanctions. All of that commentary is correct, and the numbers are genuinely alarming. But sometimes, it is worth stepping back and looking carefully at a single tree.
On March 2, 2026, this blog’s modest proposal on the abuse of AI issues mentioned the LeDoux v. Outliers, Inc. case. No. 3:24-CV-05808-TMC, 2026 WL 291023 (W.D. Wash. Feb. 4, 2026). This case is one of those cases that warrants a closer read. Not because it involves a single rogue citation slipping through the cracks, but because of the sheer breadth, variety, and audacity of the errors catalogued in the court’s opinion. Reading the opinion carefully is a useful exercise because it illustrates just how many ways AI hallucinations can infect a case, and just how difficult the cleanup can be once the infection has spread.
The LeDoux case is a case involving allegedly defective nootropic (cognition improving) supplements. The plaintiff, a medical nurse officer in the United States Army, sued the defendants, various entities and individuals involved in the design, manufacture, and sale of the nootropic supplements at issue, for negligence, unfair trade practices, failure to warn, design and manufacturing defect, breach of warranty, and misrepresentation and fraud in the Western District of Washington. Plaintiff alleged that defendants’ nootropic supplements were adulterated with amphetamines, causing her to test positive on a military drug screen, which triggered a felony court-martial prosecution and resulted in the destruction of her military career, loss of her pensions and benefits, and severe emotional harm. The legal and factual issues, however, were quickly overshadowed by something else entirely.
Defendants identified multiple citations in plaintiff’s proposed amended complaint, Daubert motion, motion to strike affirmative defenses, expert evidence, and response to defendants’ motion to strike experts, which they believed to be false and hallucinated by AI. All told, these AI hallucinations spanned more than five separate filings. And the defense did not just flag the problem in a footnote. They did the work and provided various tables for the court describing the faulty citations. That kind of methodical, document-by-document, citation-by-citation opposition work is what turned a case about supplements into a case about professional misconduct.
The breadth of what defense counsel uncovered is worth going through carefully, because it illustrates how thoroughly AI-generated sloppiness can permeate a plaintiff’s entire case. Plaintiff’s proposed amended complaint quoted text from a statute, RCW 7.72.010(2), that did not exist. For example, plaintiff alleged that a defendant “performed functions that constitute manufacturing activity within the meaning of RCW 7.72, including ‘supplying a component part,’” but the term “supplying a component part” did not exist in RCW 7.72, and the definition of “manufacturer” in RCW 7.72.010(2) did not include someone who “supplies” a component part. Id. at *1.
Plaintiff also submitted over a dozen quotes from various expert reports that did not exist. For example, plaintiff cited the deposition of one of her experts to argue that dietary supplements are uniquely prone to “post-production degradation of stimulant compounds,” but that quote did not appear anywhere in the deposition transcript. Id. at *2. Plaintiff also cited nonexistent cases and included quotes from other cases that did not exist in the cited cases. Each of these was caught by defense counsel checking the actual sources.
Perhaps the most strategically significant finding for defense purposes involved hallucinations in plaintiff’s expert reports (something we’ve also addressed). One expert report cited several articles that did not exist or were attributed to the wrong author, and another expert report cited the same articles and included the same citation errors. The court found this “suggest[ed] [p]laintiff’s counsel may have been so involved with the drafting of expert reports as to ‘jeopardize the experts’ independence.’” Id. (citation omitted).
Read More
AI Hallucinations in Court: A Case Study in How Bad It Can Get
