According to the Ninth Circuit opinion in LNU v. Blanche (No. 24-4790), authored by Judges Richard Paez, Carlos Bea, and Danielle Forrest, the court issued a disciplinary order after finding that attorneys Mike Singh Sethi and William Rounds filed briefs containing nonexistent cases, misattributed quotations, and other misrepresentations, some of which the attorneys attributed to typing errors. The opinion defines two relevant types of generative AI hallucinations, "fabrications" and "inaccuracies," and states: "be aware of the risks of overreliance on generative AI, read everything cited in a court filing, and disclose quickly and transparently generative AI hallucinations that are inadvertently included in court filings." The court ordered the attorneys to show cause and imposed discipline, per the opinion and reporting by Reason and the court PDF.
What happened
Per the Ninth Circuit opinion in LNU v. Blanche (No. 24-4790), authored by Judge Richard Paez and joined by Carlos Bea and Danielle Forrest, the court issued a disciplinary order addressing briefs filed by attorneys Mike Singh Sethi and William Rounds that contained "nonexistent cases, misattributed quotations, and gross misrepresentations of real cases," according to the opinion text reproduced in the court PDF and summarized by Reason. The opinion recounts that the attorneys claimed typographical errors and repeatedly denied that generative AI produced the errors. The court ordered the attorneys to show cause why they should not be sanctioned, suspended, or disbarred, and it ultimately imposed discipline.
Technical details
The opinion classifies two categories of generative AI mistakes as most relevant: "fabrications", where the tool creates cases or quotations that do not exist, and "inaccuracies", where the tool cites real authorities but provides legally or factually incorrect statements relative to those authorities. The opinion warns that inaccuracies can be subtler and more dangerous because they may survive cursory checks and require close analysis of cited sources. The court instructs attorneys that, when they learn of any error in a filing, including generative AI hallucinations, they should immediately alert the court, per the court document.
Industry context
Editorial analysis: Courts, law firms, and legal-tech teams have been tracking generative AI hallucinations as a practical and ethical risk for legal writing and research. Public court discipline for filing fabricated or inaccurate authorities elevates the operational stakes for lawyers using generative tools. Observers have noted that hallucinations that mimic plausible citations are harder to detect than obvious fabrications, which raises risk management and verification burdens.
What to watch
For practitioners: monitor whether other courts adopt similar admonitions or formal filing rules concerning disclosure of AI assistance. Also watch professional-responsibility guidance from bar associations and how e-discovery and citation-checking workflows adapt. For legal-tech teams, improvements in citation groundedness and automated verification tooling will be an immediate area of focus given the court's explicit concern about undetected inaccuracies.
Implications
Editorial analysis: The decision links generative-AI hallucinations directly to professional-discipline outcomes in at least one federal circuit. That establishes a clearer precedent that reliance on generative outputs without verification can have career-level consequences for practitioners, and it highlights the importance of workflow controls that surface and correct both fabricated and subtle inaccurate citations.
Scoring Rationale #
A federal appellate court imposing discipline over filings that include generative-AI errors is a notable legal precedent with direct implications for practitioners and legal-tech tooling. The story is highly relevant to legal and verification workflows but has narrower technical impact for core ML research.
Practice interview problems based on real data
1,500+ SQL & Python problems across 15 industry datasets — the exact type of data you work with.