LeDoux v. Outliers, Inc.
U.S. District Court, Western District of Washington · W.D. Wash. · Washington bar guidance
Conduct
Counsel used ChatGPT and Claude to draft filings and an expert-report citation table, and filed dozens of false citations and quotations without checking them.
Consequence
Here the court imposed a $3,000 personal Rule 11 sanction, a citation certification on later briefing, and client notice, and later excluded an expert report over hallucinated citations.
Lesson
Here the court found that failing to act after opposing counsel flagged the citations, and a flawed correction, supported a finding tantamount to bad faith.
Verified September 28, 2026
- Citation
- LeDoux v. Outliers, Inc., No. 3:24-cv-05808-TMC (W.D. Wash. July 24, 2026) (Cartwright, J.) (Order on Sanctions, Dkt. 265); see id. (W.D. Wash. Feb. 4, 2026) (Order to Show Cause, Dkt. 209)
- Decided
- July 24, 2026
Summary
In a consumer suit over nootropic supplements, defendants flagged at a December 2, 2025 hearing and in later filings that plaintiff Joann LeDoux's briefs and expert evidence contained citations they believed were "hallucinated" by artificial intelligence. On February 4, 2026 (Dkt. 209), Judge Tiffany M. Cartwright found that "Plaintiff has submitted dozens of inaccurate factual and legal citations across at least five different filings," including a citation to "Berg v. Chevrolet Motor Div., 84 Wn.2d 102," which "does not exist," and quotations that do not appear in a cited decision. The court found counsel's explanations and corrections "incomplete at best and misleading at worst"; a "corrected" Daubert motion replaced some false quotations with real ones but "completely changed their meaning." Two expert reports cited the same nonexistent or misattributed academic articles. The court ordered counsel to show cause why she should not be sanctioned under Rule 11, by a sworn declaration detailing all incorrect citations, giving "a thorough explanation for how the incorrect citations discussed in this order were found or generated," and naming any AI programs used. In her declaration, Jocelyn Stewart stated that she used ChatGPT and Claude, that they generated the false citations, and that she "did not verify the output against the source before filing."
- AI tool:
- ChatGPT and Claude (per counsel's sworn declaration, as recounted in the July 24, 2026 sanctions order)
- Sanction amount:
- $3,000 personal sanction on attorney Jocelyn Stewart, payable to the Clerk; certification of verified citations on all later briefing in the case; copy of the order to the client
What sanction did the court impose?
On July 24, 2026 (Dkt. 265), the court sanctioned Stewart sua sponte under Rule 11. It found her conduct "exceed[ed] mere ignorance or negligence" and was "tantamount to bad faith," pointing to the volume of errors, her involvement in an AI-generated citation table given to two experts, and her failure to act after defendants flagged the problems, including one more hallucinated citation filed on December 18, 2025. It credited her full admission of fault but said her concessions "cannot replace the time and resources lost." Stewart was personally sanctioned $3,000, must certify that all citations are verified in all later briefing in the case, and had to give the order to her client and certify that she had done so. The show-cause order was discharged; the order makes no bar referral. On August 18, 2026 (Dkt. 269), the court excluded plaintiff's expert Dr. Gabriel Holguin, finding that "the hallucinated citations undermine Dr. Holguin's entire report," excluded another expert on other grounds, granted summary judgment, and dismissed the remaining claims with prejudice.
Why does LeDoux v. Outliers, Inc. matter for law firms using AI?
LeDoux is a volume case: dozens of defective citations and quotations across four briefs and two expert reports, followed by explanations and corrections the court called “incomplete at best and misleading at worst” (Dkt. 209) and, in the sanctions order, “incomplete and confusing” (Dkt. 265). Judge Cartwright’s February 2026 show-cause order required counsel to detail all incorrect citations, give “a thorough explanation for how the incorrect citations discussed in this order were found or generated,” and name any AI programs used. Counsel’s sworn declaration named ChatGPT and Claude and conceded that she did not conduct a reasonable inquiry before signing, explaining that she believed the tools, working from documents she supplied in a “closed system,” would produce accurate output.
The sanctions order turns on the Ninth Circuit’s “akin to contempt” standard for sua sponte Rule 11 sanctions. The court found counsel’s conduct tantamount to bad faith for three stated reasons: the volume and obviousness of the errors, which showed the filings were not subjected to meaningful scrutiny; the AI-generated citation table she supplied to two experts, which the court said suggested involvement that could jeopardize their independence; and her failure to act after defendants flagged the problems in December 2025, including filing another hallucinated citation and an incomplete “corrected” motion. Her full admission in response to the show-cause order was credited but did not avoid a sanction; the court imposed $3,000, a continuing citation certification, and client notice, and did not refer her to a disciplinary body.
The case also shows the downstream cost to the client’s evidence. In August 2026 the court excluded another plaintiff expert because hallucinated citations in withdrawn portions of his report undermined “the overall credibility” of the whole report, then granted summary judgment and dismissed the remaining claims with prejudice on damages grounds. Firms documenting compliance may wish to consider how AI-assisted preparation of expert materials is reviewed, since here the errors affected both counsel’s sanctions exposure and the admissibility of the client’s expert evidence.
Implications for your firm
Operational steps a firm reading this case may wish to consider documenting. Strategic and rule-application calls belong to your firm's attorneys.
- Verify citations in materials counsel prepares for experts; here an AI-generated citation table given to two experts carried the same errors into both reports, and a third expert's report was later excluded.
- Review every AI-assisted filing once opposing counsel flags a false citation; here the court faulted counsel for filing another hallucinated citation two weeks after the problem was raised.
- Document corrections accurately; here a "corrected" Daubert motion that changed the meaning of the argument, and left some errors in place, weighed toward bad faith.