Ruiz v. Magellan Financial & Insurance Services
U.S. District Court, District of Arizona · D. Ariz. · Arizona bar guidance
Conduct
Counsel used Claude AI to draft a fee-motion response with two false quotations; the court then found similar apparent errors in three of her other briefs.
Consequence
Here the court imposed no sanction yet; it ordered a supplemental memorandum addressing the additional errors before deciding how to proceed.
Lesson
Here the court treated checking an AI-generated list of cited cases, rather than the brief itself, as a verification method with an obvious flaw.
Verified September 29, 2026
- Citation
- Ruiz v. Magellan Financial & Insurance Services, No. CV-23-02090-PHX-DWL, Order (D. Ariz. July 28, 2026), ECF No. 172
- Decided
- July 28, 2026
Summary
On July 8, 2026, the court ordered plaintiff's counsel Elizabeth Tate to show cause why she should not be sanctioned or disciplined for two false quotations from Ninth Circuit cases in the plaintiff's response to the defendant's motion for attorneys' fees. Her response, as quoted in the order, said she used Claude AI to research and prepare the brief and asked Claude to list the cases it had cited so she could review them on Lexis; she read the cases, fixed other misquotations Claude produced, but "mistakenly neglected to fix" the two at issue. She called her AI use responsible and the errors isolated. The court said it appeared that her method checked Claude's list of cases rather than the brief itself, spot-checked her other filings, and found that at least three other briefs appeared to contain similar errors: a motion in limine citing Ansell v. Green Acres Contracting Co. for excluding "good acts" evidence, though, as the court agreed with the defendant, Ansell affirmed admission of evidence of favorable treatment of another older worker; quotations attributed to United States v. Antonakeas and United States v. Osazuwa that do not appear in those opinions; and a purported quotation of 28 U.S.C. § 1924 that the court called "a fabricated paraphrase of the statutory language masquerading as a direct quotation."
- AI tool:
- Claude AI (counsel's show-cause response, as quoted in the order, says she used Claude AI to research and prepare the brief; the court says further errors in three other briefs "may also be attributable to the misuse of generative AI")
How did the court address the AI-related conduct?
The court did not rule on sanctions in this order. It said it appeared that the additional examples "may also be attributable to the misuse of generative AI" and, if so, were "difficult to reconcile with the notion" that the two false quotations in the fees response were "an isolated mistake stemming from the otherwise responsible use of AI." Because counsel had not had a chance to address the additional examples, the court ordered her to file, within 14 days, a supplemental memorandum of no more than 10 pages addressing the issues the order raised. The court also rejected counsel's description, in a July 9, 2026 notice of errata, of the § 1924 quotation as merely "incomplete," calling that "benign characterization" inaccurate. In a footnote it acknowledged that Antonakeas and Osazuwa contain passages that broadly support the points counsel was making, but cited the Ninth Circuit's statement in Lnu v. Blanche that it "makes no difference" that misquoted language is generally supported by the cited case. The later August 24, 2026 order resolving the show-cause proceeding, which issued a formal public reprimand, has its own entry.
Why does Ruiz v. Magellan Financial & Insurance Services matter for law firms using AI?
This July 28 order is the middle step in the Ruiz show-cause proceeding. The court’s July 8 order to show cause addressed two false quotations in a single brief. Counsel’s response, as the court quotes it, described a workflow in which Claude AI drafted the brief and then produced a list of the cases it had cited, which she requested so she could review them on Lexis. She said she had caught and fixed several other misquotations Claude produced but “mistakenly neglected to fix” the two at issue, and she asked the court to forgo sanctions or, at most, to issue a reprimand or admonition.
The court said that counsel’s “response to the OSC raises additional questions that require further development.” It identified two problems with the workflow: checking the AI’s own list of cases provides no way to confirm the list is accurate, and delegating the entire drafting process “(other than a hasty cite-check at the end)” to a generative AI program is, in the court’s words, hard to characterize “as a responsible use of AI.” Because counsel had framed the two quotations as isolated, the court spot-checked her other filings and found that “it appears that at least three other briefs contain similar errors,” including a statutory quotation that her own errata notice had described as merely “incomplete.”
The order makes no finding that AI caused the additional errors and imposes no sanction. The court said it appeared that the additional examples “may also be attributable to the misuse of generative AI” and gave counsel 14 days to address them. The court resolved the proceeding on August 24, 2026 with a formal public reprimand; see Ruiz v. Magellan Financial & Insurance Services. Firms documenting compliance may wish to consider that, in this case, counsel’s own description of her AI verification process prompted the court’s wider review of her filings.
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 and quotations in the text of the brief as filed. Here the court said checking a Claude-generated list of cases, instead of the brief line by line, 'does not include a mechanism for verifying the accuracy of the list generated by Claude.'
- Review how much drafting is delegated to an AI tool. The court said it did not strike it as responsible to let an AI program draft an entire brief and submit it 'without significant attorney reworking' as the attorney's own work product, citing the State Bar of Arizona's April 2025 guidance.
- Document an accurate description of any error in a notice of errata. Here the court rejected counsel's description of a fabricated statutory quotation as 'incomplete.'
- Consider that a court may review a lawyer's other filings in the case after one AI error surfaces; here that review turned up three more briefs with apparent errors.