Akerlund v. Atlas Air: a federal appeals court refers a former Florida lawmaker to its lawyer-conduct committee over AI-hallucinated citations, filed twice

The problem

A group of commercial-aviation employees sued Atlas Air, Inc. and Flight Services International, LLC over the companies’ COVID-19-era mask, testing, and vaccination policies, and the district court dismissed their third amended complaint (source). They appealed to the U.S. Court of Appeals for the Eleventh Circuit, which affirmed the dismissal but had a second matter to address (source). As the opinion opens: “We are more candid than usual in this assessment because the plaintiffs’ counsel Anthony F. Sabatini has not been candid with us. Sabatini filed multiple briefs replete with fake and hallucinated citations” (source). Sabatini is a former member of the Florida House of Representatives (source).

What was built

This is not a technology deployment; it is an AI-governance outcome, and the court found the fabrication was not a stray error but a pattern produced by unverified AI drafting (source). Sabatini’s “opening brief relies on at least eight such cases, including one purportedly decided by this Court” (source). After the defendants identified the problem, he acknowledged in a proposed reply brief that the citations were “erroneous or unverifiable” and sought to withdraw eight cases (source). The court noted that “the eight cases Sabatini ‘withdrew’ did not match a single one of the eight hallucinated cases in his opening brief,” and that “not only were they not the right cases, all eight were also hallucinated” (source). It concluded, more broadly, that “By outsourcing his legal work to an AI algorithm, Sabatini violated his ethical duties to both his clients and this Court,” adding that “Whatever the merits of artificial intelligence, it is no substitute for actual intelligence” (source). No specific AI product is named in the opinion (source).

The outcome

The scale of the fabrication: at least eight nonexistent cases in the opening brief, then eight more in the proposed reply, all hallucinated. On July 10, 2026, in a published opinion by Circuit Judge Grant, joined by Circuit Judges Rosenbaum and Kidd, the Eleventh Circuit affirmed the dismissal for lack of personal jurisdiction and failure to state a claim (source).

The penalty is non-monetary: a disciplinary referral, not a fine. The court said it would, “through the Chief Judge,” refer the matter to the Committee on Lawyer Qualifications and Conduct, and it imposed no dollar sanction (source). Two mutually independent newsrooms corroborate the finding and the referral: Reason/The Volokh Conspiracy, which reproduces the opinion, and the ABA Journal, which reported that the Eleventh Circuit reprimanded “a Florida lawyer and county official for filing briefs that included ‘fake and hallucinated’ artificial intelligence-generated information,” that Sabatini “produced multiple briefs containing nonexistent citations,” and that “The 11th Circuit will refer Sabatini to the court’s Committee on Lawyer Qualifications and Conduct” (source).

How this was verified

The origin is a published federal appellate opinion, which is self-authenticating: no client-confirmation call is booked against the court, and no green badge is claimed here. Every figure on this page was re-checked live on 2026-08-18 against the primary opinion (CourtListener’s copy of the Eleventh Circuit slip opinion, No. 24-11033) and two mutually independent newsrooms, Reason/The Volokh Conspiracy and the ABA Journal. There is no monetary figure, because the sanction is a disciplinary referral rather than a fine; the “at least eight plus eight” count and the referral both re-verified without change. No AI product is named in the record, and no claim here asserts one. This story remains pending: before any human sign-off, the record should be checked for rehearing or en banc review of the July 10, 2026 opinion (Eleventh Circuit docket, No. 24-11033) and for the separate Chief-Judge order effecting the referral.

Related case files

• The Ninth Circuit reached a comparable posture in Lnu v. Blanche, where a federal appeals panel sanctioned counsel for AI-fabricated authorities and referred them to a bar disciplinary body, the same referral mechanism used here.
• Gauthier v. Goodyear is an early archetype of the pattern: hallucinated citations drew a sanction plus mandatory generative-AI legal education, showing how courts began responding before referrals became common.
• United States v. Hayes marks the other end of the range, a court that imposed a modest fine but expressly declined to find AI use, which is the finding Akerlund makes explicitly.

Sources

Tier 1 is the party’s own record or a court document; Tier 2 is independent reputable press naming the parties. Figures on this page were checked live on 2026-08-18.

1. U.S. Court of Appeals for the Eleventh Circuit, “Akerlund v. Atlas Air, Inc.,” No. 24-11033 (FOR PUBLICATION), 2026-07-10 (Tier 1, primary appellate opinion; copy hosted by CourtListener). https://storage.courtlistener.com/pdf/2026/07/10/estate_of_lane_caviness_v._atlas_air_inc..pdf
2. Reason / The Volokh Conspiracy, “Florida Politician (Anthony F. Sabatini) Faulted for AI Hallucinations in Briefs,” 2026-07-11 (Tier 2, reproduces the opinion and quotes it at length). https://reason.com/volokh/2026/07/11/florida-politician-anthony-f-sabatini-faulted-for-ai-hallucinations-in-briefs/
3. ABA Journal, “Lawyer’s use of ‘fake and hallucinated’ citations gets reprimand from U.S. appeals court,” 2026 (Tier 2, independent narration of the ruling and referral). https://www.abajournal.com/news/article/lawyers-use-of-fake-and-hallucinated-case-gets-reprimand-from-us-appeals-court