Lichfield v. Kubler: 10th Circuit finds AI-fabricated citations but declines to sanction

The problem

Attorneys are increasingly filing briefs whose case citations were produced by generative-AI tools and never verified, and when those tools “hallucinate” they generate citations that look real but refer to cases that do not exist or attach quotations to real cases that never contained them. Federal courts have begun policing the practice — sometimes with money sanctions and disciplinary referrals, sometimes with a published warning — and Lichfield v. Kubler is an example of the latter, a case the Volokh Conspiracy’s weekly appellate roundup flagged as one that “has it all” (source).

What was built

On appeal from the dismissal of Narvin Lichfield’s defamation suit against filmmaker Katherine Kubler and Netflix over the 2024 documentary The Program: Cons, Cults and Kidnapping, the Tenth Circuit found that the appellant’s opening brief contained fabricated or hallucinated case citations (source). The court did not name the specific tool but wrote that it “assume[d] the use of generative artificial intelligence led to the misrepresentations of legal authority” (source). Because the brief cited authorities that did not exist, the court held that “[t]he Opening Brief in this case therefore failed to fully comply with Rule 46.5” and that “[t]he filing of this non-compliant brief is conduct sanctionable at our discretion” (source). Counsel “filed the next day a Notice of Errata, identifying the specific errors in the opening brief and providing actual, existing authority” (source).

The AI attribution here is the court’s own assumption from the nature of the errors, not an admission by counsel and not a factual finding that a particular tool was used, and that is the honest limit of the AI angle in this record (source).

The outcome

Unlike the same court’s earlier decision in Amarsingh v. Frontier Airlines — where it ordered a $1,000 payment and a disciplinary referral — the panel here declined to sanction, reasoning that “the brief here had significantly fewer errors than in Amarsingh” and concluding that “[t]he missteps in this case do not merit sanctions” (source). The court nonetheless warned that “[s]ubmission of a brief containing fabricated or hallucinated citations is a divergence from the standards expected of attorneys appearing before this court” (source). On the merits it affirmed the dismissal, holding that “[u]nder Utah defamation law, Utah constitutional law, and the First Amendment,” none of the challenged statements was capable of defamatory meaning (source). The ABA Journal’s Bryan Garner reported the same AI paragraph as an instance of a circuit court expressing muted exasperation, describing an opinion that “concluded … with a paragraph that probably mortified counsel” (source), and Courthouse News reported the underlying First Amendment dismissal firsthand, noting the panel “upheld the dismissal of Narvin Lichfield[‘s]” suit (source).

A note on the weakest load-bearing source: every critical fact above is anchored to the Tier-1 published opinion itself; the independent read-throughs (ABA Journal, the Volokh Conspiracy, Courthouse News) are Tier-2 secondary commentary, not primary filings, and they corroborate the AI angle and the dismissal rather than establishing them. The one figure that could mislead — the $1,000 — belongs to the earlier Amarsingh case, cited here only as contrast; Lichfield itself imposed no monetary sanction and made no disciplinary referral (source).

How this was verified

• Method: primary-document verification against a signed, PUBLISHED federal appellate opinion (10th Cir., No. 25-4135), with every quoted line grep-confirmed verbatim in the extracted opinion text and the AI angle corroborated by two independent newsrooms.
• Primary: the opinion PDF is byte-tied to the CourtListener durable store — local capture sha1 21818ef83bd54a56aca6ae9ae804c8fdc104ef6c (422,528 bytes), identical to the live re-download.
• Date verified: 2026-08-19 (research settled 2026-07-31; opinion filed 2026-07-27).
• Status: court-origin honest-negative — the facts come from the opinion, not any party’s self-report, so no client-confirmation call is required. The remaining human step before a green badge is editorial: confirm on the live Tenth Circuit docket (No. 25-4135) that no later rehearing or modification has disturbed the no-sanction disposition.

Related case files

• Amarsingh v. Frontier Airlines — the same Tenth Circuit’s $1,000 sanction that this opinion expressly distinguishes
• United States v. Hayes — another federal case where the court sanctioned a fictitious citation but declined to find AI use
• Whiting v. City of Athens — a Sixth Circuit companion where fabricated citations drew a $15,000 punitive sanction instead of a pass

Sources

1. U.S. Court of Appeals for the Tenth Circuit · Lichfield v. Kubler; Netflix, Inc., No. 25-4135 (published opinion, Federico, J.) · 2026-07-27 · https://www.ca10.uscourts.gov/sites/ca10/files/opinions/010111477318.pdf — Tier 1 (signed, published federal appellate opinion; byte-tied to CourtListener store https://storage.courtlistener.com/pdf/2026/07/27/lichfield_v._kubler.pdf)
2. ABA Journal (Bryan Garner) · “The discipline of getting caught” · 2026-07-30 · https://www.abajournal.com/columns/article/the-discipline-of-getting-caught — Tier 2 (independent reputable press; firsthand read of the AI paragraph)
3. The Volokh Conspiracy / Reason.com (John Ross, “Short Circuit”) · weekly compendium of federal appellate rulings · 2026-07-31 · https://reason.com/volokh/2026/07/31/short-circuit-an-inexhaustive-weekly-compendium-of-rulings-from-the-federal-courts-of-appeal-70/ — Tier 2 (second independent newsroom reading the opinion firsthand)
4. Courthouse News Service · “Netflix dodges defamation suit over troubled-teen documentary ‘The Program’” · 2026-07-27 · https://www.courthousenews.com/netflix-dodges-defamation-suit-over-troubled-teen-documentary-the-program/ — Tier 2 (independent press; firsthand on the First Amendment dismissal)