AI sanctions case tracker

Lnu v. Blanche, No. 24-4790

U.S. Court of Appeals for the Ninth Circuit · June 3, 2026 · Verified against the court’s published document on July 1, 2026.

Published disciplinary order sanctioning Mike Singh Sethi and William Rounds for briefs filed "with multiple nonexistent cases, misattributed quotations, and gross misrepresentations of real cases," followed by repeated denials of AI use. Each attorney was fined $2,500 and suspended from practice before the Ninth Circuit for six months, with a two-year requirement that future filings disclose under penalty of perjury whether generative AI was used, name the tool, and certify personal review, and with the order referred to the State Bar of California. The court wrote that the rules "are not violated at the point of research and drafting, but at the point of signing and filing," and that the gravity of the discipline is owed to the repeated failure of candor.

• Sanction: Monetary sanction; six-month suspension from practice before the Ninth Circuit; two-year AI-use certification condition on all filings; referral to the State Bar of California
• Amount: $2,500 each (two attorneys)
• Authority: FRAP 46(b); 9th Cir. R. 46-2(a)
• AI tool alleged (per the primary document): None named in the order; the attorneys conceded it was "more likely than not" unauthorized generative-AI use by unlicensed brief writers

Davis v. Marion County Superior Court Juvenile Detention Center, No. 1:24-cv-01918

U.S. District Court for the Southern District of Indiana · May 20, 2026 · Verified against the court’s published document on July 21, 2026.

The court imposed a $2,000 Rule 11 sanction on attorney Tae Sture for fabricated citations it inferred came from generative AI, though counsel claimed the tool was Fastcase. The frequently cited "2025 WL 2502308 (Sept. 2, 2025)" is the magistrate's report recommending $7,500; the district judge later reduced the sanction to $2,000 and rejected the recommended disciplinary referral.

• Sanction: Monetary sanction against counsel personally (the district judge reduced the magistrate's recommended amount and declined the recommended professional-conduct referral)
• Amount: $2,000 (against attorney Tae Sture)
• Authority: Fed. R. Civ. P. 11(b)
• AI tool alleged (per the primary document): Unnamed generative AI — court inference only; counsel attributed the citations to Fastcase, and the court made no express finding that AI was used

Williams v. Honl, No. A186656

Oregon Court of Appeals · April 22, 2026 · Verified against the court’s published document on July 21, 2026.

Reported at 348 Or App 505. The court ordered appellant's attorneys to pay $8,044.25 in respondent's attorney fees for a brief in which six of nine cited cases were fabricated or misused by an unidentified generative-AI tool, treating the uncorrected record as a candor violation.

• Sanction: Award of respondent's attorney fees against appellant's attorneys
• Amount: $8,044.25 (against appellant's attorneys)
• Authority: ORAP 1.40 and ORCP 17 (fees via ORCP 17 D(4)); RPC 3.3 (candor)
• AI tool alleged (per the primary document): Unnamed generative AI — "Appellant's attorney did not identify what generative artificial intelligence product she used"; six of nine cited cases were fabricated or misused

United States v. Farris, No. 25-5623

U.S. Court of Appeals for the Sixth Circuit · April 3, 2026 · Verified against the court’s published document on July 21, 2026.

The Sixth Circuit stripped attorney Steven N. Howe of his CJA compensation, removed him from the representation, and made three disciplinary referrals for a brief whose quotations were fabricated by Westlaw's CoCounsel AI. Cited by docket: secondary sources circulated "2026 WL 915082," but the correct Westlaw cite is 2026 WL 710568.

• Sanction: Forfeiture of Criminal Justice Act compensation for the appeal; removal from the representation; disciplinary referrals (Sixth Circuit and E.D. Ky. Chief Judges; Kentucky Bar)
• Amount: None fixed (CJA fee forfeiture plus referrals)
• Authority: 18 U.S.C. § 3006A; 6th Cir. Local Rule 46; CJA Plan; inherent authority
• AI tool alleged (per the primary document): Named — Westlaw's CoCounsel ("Westlaw's internal artificial-intelligence platform"); the cited cases were real but their quotations were fabricated or misstated

Prososki v. Regan, No. S-25-295

Nebraska Supreme Court · March 20, 2026 · Verified against the court’s published document on July 19, 2026.

Per curiam opinion (321 Neb. 38) striking an appellate brief that contained "numerous citations to fake cases, real cases with fake quotations, real cases with mischaracterized holdings, and Nebraska statutes and court rules with fake quotations," dismissing the appeal, and referring counsel to the Counsel for Discipline. The opinion prints a chart of roughly twenty problematic citations, including a fictitious "Kennedy v. Kennedy" cited and quoted five times, and holds that submitting fictitious authority — "whether through generative AI or not" — is resolved by existing candor, competence, and frivolous-claim rules. The court reviewed the record for plain error despite striking the brief, found none, and affirmed on the cross-appeal.

• Sanction: Appellant's brief stricken; appeal dismissed; counsel referred to the Counsel for Discipline of the Nebraska Supreme Court
• Amount: None (non-monetary; the court noted its ruling does not preclude the appellee from seeking attorney fees)
• Authority: Inherent authority; Neb. Ct. R. App. P. § 2-109(D)(1)(e) briefing mandate; referral for investigation of potential violations of Neb. Ct. R. of Prof. Cond. §§ 3-503.3(a)(1) (candor), 3-501.1 (competence), 3-501.3 (diligence), 3-503.1 (meritorious claims), and 3-508.4(c) (misconduct)
• AI tool alleged (per the primary document): None found — the appellee asserted the fabrications were the product of generative AI; counsel denied inventing cases or relying on AI; the court expressly declined to resolve the question, holding that "[r]egardless of whether AI was used" the analysis is the same

Doiban v. Oregon Liquor and Cannabis Commission, No. A181982

Oregon Court of Appeals · March 18, 2026 · Verified against the court’s published document on July 21, 2026.

Reported at 347 Or App 742. Oregon's largest appellate AI-fabrication sanction to date: $10,000 against attorney William L. Ghiorso for at least 15 fabricated citations and 9 false quotations, expressly applying and capping the fee schedule the court originated in Ringo. Decided March 18, 2026 — not December 2025, a date some secondary reports carry over from Ringo.

• Sanction: Monetary sanction against counsel, capped below the amount the court's per-infraction formula would produce
• Amount: $10,000 (against attorney William L. Ghiorso; capped below the roughly $16,500 that 15 fabricated citations and 9 false quotations would yield under the Ringo formula)
• Authority: ORCP 17 C(3), via ORAP 1.40(4)
• AI tool alleged (per the primary document): Generative AI (via search-engine AI) — counsel's staff relied on search engines including Google, whose AI-generated responses affirmed the nonexistent cases; the court tied the fabrications to generative AI's "tendency to fabricate" and applied its Ringo framework

Whiting v. City of Athens, Tenn., Nos. 24-5918/5919, 25-5424

U.S. Court of Appeals for the Sixth Circuit · March 13, 2026 · Verified against the court’s published document on July 1, 2026.

Published opinion sanctioning Van R. Irion and Russ Egli for briefs containing "over two dozen fake citations and misrepresentations of fact" — the court's conservative estimate, itemized in an appendix — across three consolidated appeals. This is a fabricated-citations case, not a confirmed AI case: the court made no express AI finding because the attorneys did not respond to the show-cause directive asking whether AI was used.

• Sanction: Punitive monetary sanction payable to the court registry; joint and several liability for appellees' full reasonable attorneys' fees on appeal; double costs under 28 U.S.C. § 1920; referral to the chief judge to consider disciplinary proceedings under 6th Cir. Local Rule 46
• Amount: $15,000 each (two attorneys), plus fees and double costs
• Authority: FRAP 38 (appeal frivolous as argued); inherent authority (bad faith)
• AI tool alleged (per the primary document): None — the court made no express finding that generative AI produced the fabrications; the attorneys never answered the show-cause order's question whether they used AI

Fletcher v. Experian Information Solutions, Inc., No. 25-20086

U.S. Court of Appeals for the Fifth Circuit · February 18, 2026 · Verified against the court’s published document on July 21, 2026.

Published order sanctioning attorney Heather Hersh $2,500 after the Fifth Circuit found her reply brief contained unsupported and fabricated citations it attributed to generative AI. Cited by docket and date: the reporter and Westlaw pagination for this opinion were not independently confirmable from free sources.

• Sanction: Monetary sanction against counsel, payable to the court
• Amount: $2,500 (against attorney Heather Hersh)
• Authority: FRAP 46(c); inherent authority
• AI tool alleged (per the primary document): Found (unnamed) — the court found counsel "used artificial generative intelligence to draft a substantial portion — if not all — of her reply brief"; no specific tool named

Powell v. Employment Department, No. A185894

Oregon Court of Appeals · February 11, 2026 · Verified against the court’s published document on July 21, 2026.

Reported at 347 Or App 55. The Oregon Court of Appeals fined a self-represented petitioner $500 for citing fabricated cases he conceded came from online search engines and AI research sites — one of the few sanctions in this dataset against a pro se party rather than an attorney.

• Sanction: Monetary sanction against a self-represented petitioner, payable to the OJD Appellate Court Services Division
• Amount: $500 (against self-represented petitioner Keith Powell)
• Authority: ORCP 17 C(1) and D(1), via ORAP 1.40(4)
• AI tool alleged (per the primary document): Conceded (unnamed) — the petitioner attributed the fabricated authority to "online search engines and AI-assisted legal research sites"

Ringo v. Colquhoun Design Studio, LLC, No. A186670

Oregon Court of Appeals · December 3, 2025 · Verified against the court’s published document on July 21, 2026.

Reported at 345 Or App 301. This decision originates Oregon's per-infraction fee schedule for AI-fabricated authority ($500 per fabricated citation, $1,000 per false quotation), sanctioning appellee's counsel Gabriel A. Watson $2,000. Some secondary trackers caption this matter by the attorney's name, Watson; it is a single case, not two.

• Sanction: Monetary sanction against appellee's counsel, payable to the OJD Appellate Court Services Division
• Amount: $2,000 ($500 per fabricated citation x 2, plus $1,000 per false quotation x 1), against attorney Gabriel A. Watson
• Authority: ORCP 17 C(3), via ORAP 1.40(4)
• AI tool alleged (per the primary document): Unnamed generative AI — court-inferred; the court "recognize[d] ... the fabricated law in respondents' brief likely resulted from the use of artificial intelligence," though counsel did not directly admit it

In re Kheir (Kheir v. Titan Team LLC), Adv. No. 25-3033

U.S. Bankruptcy Court for the Southern District of Texas · November 4, 2025 · Verified against the court’s published document on July 21, 2026.

Chief Bankruptcy Judge Rodriguez ordered attorney Derrick D. King to reimburse the opposing party's fees and costs for a brief containing 32 fabricated or misrepresented AI-generated citations. Cited by docket: the "674 B.R. 631" reporter pincite carried in secondary sources appears in no primary or secondary record and is not used here.

• Sanction: Attorney ordered to reimburse the opposing party's reasonable attorney fees and costs (amount fixed by later fee application)
• Amount: Opposing party's reasonable attorney fees and costs (amount set by fee application)
• Authority: Rule 11; inherent authority
• AI tool alleged (per the primary document): Generative-AI hallucinations — the court's admonition references ChatGPT ("No lawyer should be using ChatGPT or any other generative AI ...") but did not affirmatively find it as counsel's specific source

In re: Sanctions Order of Kerry Kenney, No. 25-C-389

Louisiana Court of Appeal, Fifth Circuit · October 23, 2025 · Verified against the court’s published document on July 21, 2026.

On supervisory writ, the Louisiana Fifth Circuit affirmed a trial-court sanction (rendered August 11, 2025) of $1,368, three hours of ethics and AI CLE, and a disciplinary referral against attorney Connie P. Trieu for filings citing fabricated authority from ChatGPT, Copilot, and Gemini. Kerry Kenney is the plaintiff, not the sanctioned lawyer; cited by docket, as the "422 So. 3d 905" pincite carried in secondary sources is unverifiable.

• Sanction: Monetary sanction payable to defendants; three hours of ethics and AI continuing legal education; referral to the Louisiana Attorney Disciplinary Board
• Amount: $1,368 (against attorney Connie P. Trieu)
• Authority: La. Code Civ. Proc. art. 863(B)(2) & (D)
• AI tool alleged (per the primary document): Named — ChatGPT, Microsoft Copilot, and Google Gemini

Noland v. Land of the Free, L.P., No. B331918

California Court of Appeal, Second District, Division Three · September 12, 2025 · Verified against the court’s published document on July 19, 2026.

Published opinion (114 Cal.App.5th 426, Edmon, P.J.) affirming summary judgment and sanctioning appellant's counsel Amir Mostafavi $10,000 payable to the clerk after finding 21 of the 23 case quotations in the opening brief were fabrications, with more in the reply brief and at least one cited case that does not exist. Published "as a warning," with the much-quoted holding that no filing should contain any citations — "whether provided by generative AI or any other source — that the attorney responsible for submitting the pleading has not personally read and verified." Both the Ninth Circuit's Lnu order and the Sixth Circuit's Whiting opinion quote this case; sanctions ran to the court's general fund rather than opposing counsel, who never flagged the fabrications.

• Sanction: Monetary sanction payable to the clerk of the court; State Bar referral under Bus. & Prof. Code § 6086.7(a)(3); counsel ordered to serve the opinion on his client and certify service
• Amount: $10,000 (against counsel only; the court found nothing indicating the client knew of the fabrications)
• Authority: Code Civ. Proc. §§ 907, 128.7; Cal. Rules of Court, rule 8.276(a)(4); briefing violation of rule 8.204(a)(1)(B)
• AI tool alleged (per the primary document): Found and conceded — the court found the fabricated authorities "were created by generative artificial intelligence (AI) tools"; counsel admitted drafting briefs, "enhancing" them with ChatGPT, running them through other AI platforms (his account names ChatGPT, Claude, Gemini, and Grok), and not reading the enhanced briefs before filing

Johnson v. Dunn, No. 2:21-cv-1701-AMM

U.S. District Court for the Northern District of Alabama · July 23, 2025 · Verified against the court’s published document on July 21, 2026.

Reported at 792 F. Supp. 3d 1241. The court publicly reprimanded three Butler Snow attorneys, disqualified them from the case, and referred them to the Alabama State Bar after they filed ChatGPT-fabricated citations they conceded were hallucinations. The sanction was non-monetary and rested on the court's inherent authority.

• Sanction: Public reprimand (with publication); disqualification from the case; referral to the Alabama State Bar and licensing authorities
• Amount: None (non-monetary)
• Authority: Inherent authority (the court held Rule 11 inapplicable to the discovery motions at issue)
• AI tool alleged (per the primary document): Named — ChatGPT; three attorneys conceded the citations were ChatGPT hallucinations and the court so found

Wadsworth v. Walmart, Inc., No. 2:23-cv-118-KHR

U.S. District Court for the District of Wyoming · February 24, 2025 · Verified against the court’s published document on July 21, 2026.

Reported at 348 F.R.D. 489. The court sanctioned three Morgan & Morgan attorneys a total of $5,000 (and revoked one pro hac vice admission) after eight of nine cited cases proved nonexistent, generated by the firm's internal AI platform. The court specifically found the fabrications did not come from ChatGPT, and the dataset preserves that distinction.

• Sanction: Monetary sanctions payable to the court registry; one attorney's pro hac vice admission revoked
• Amount: $5,000 total — $3,000 (Ayala), $1,000 (T. Morgan), $1,000 (Goody)
• Authority: Fed. R. Civ. P. 11(b)(2); inherent authority
• AI tool alleged (per the primary document): Conceded generative AI — an internal Morgan & Morgan AI platform; the court expressly noted it "appears Plaintiffs' attorneys did not use ChatGPT." Not a ChatGPT case.

United States v. Hayes, No. 2:24-cr-0280-DJC

U.S. District Court for the Eastern District of California · January 17, 2025 · Verified against the court’s published document on July 21, 2026.

The court sanctioned an Assistant Federal Public Defender $1,500 for a fabricated case citation bearing the hallmarks of AI, expressly declining to find that AI was in fact used. Cited by docket; the "763 F. Supp. 3d 1054" reporter pincite was not independently confirmed. This corrects a roster note that had flagged Hayes as authority-only: it imposed its own sanction.

• Sanction: Monetary sanction against counsel personally; the order served on the D.C. Bar and the State Bar of California and circulated to the court's judges
• Amount: $1,500 (against Assistant Federal Public Defender Andrew Francisco)
• Authority: E.D. Cal. Local Rule 180(e); inherent authority
• AI tool alleged (per the primary document): No express AI finding — the court said the fabricated citation "has all the markings of a hallucinated case created by generative artificial intelligence (AI) tools such as ChatGPT and Google Bard" but "need not make any finding" on actual AI use

Grant v. City of Long Beach, Nos. 22-56121, 22-56143

U.S. Court of Appeals for the Ninth Circuit · March 22, 2024 · Verified against the court’s published document on July 19, 2026.

Published opinion (Desai, J., joined by H.A. Thomas, J., and Márquez, D.J.) striking an opening brief "replete with misrepresentations and fabricated case law" and dismissing the appeal. The panel documented two cited cases that "do not appear to exist" and a dozen real cases whose facts and holdings were misrepresented; at oral argument, counsel did not acknowledge the fabrications even when asked directly. The Ninth Circuit's 2026 Lnu order cites Grant as its precedent for striking a brief "composed almost entirely of fabricated or inaccurate citations." Included as a fabricated-citations sanction; the opinion makes no AI finding at all, and the dataset preserves that distinction.

• Sanction: Opening brief stricken in its entirety; appeal dismissed
• Amount: None (non-monetary)
• Authority: 9th Cir. R. 28-1 (material failure to comply with briefing rules); FRAP 28(a)(8)(A)
• AI tool alleged (per the primary document): None — the published opinion (96 F.4th 1255) never mentions artificial intelligence; it documents fabricated case law and misrepresented authorities without any finding as to their source

Kruse v. Karlen, No. ED111172

Missouri Court of Appeals, Eastern District · February 13, 2024 · Verified against the court’s published document on July 21, 2026.

Reported at 692 S.W.3d 43. The Missouri Court of Appeals dismissed a self-represented appellant's appeal in which 22 of 24 case citations were fictitious and awarded $10,000 in frivolous-appeal damages, attributing the fabricated authorities to a generative-AI tool the appellant did not identify.

• Sanction: Damages for a frivolous appeal, toward respondent's appellate attorney fees; appeal dismissed
• Amount: $10,000 (against self-represented appellant Jonathan Karlen)
• Authority: Mo. Sup. Ct. R. 84.19 (also Rules 84.04, 55.03, 84.06(c))
• AI tool alleged (per the primary document): Unnamed generative AI — the court attributed 22 fictitious citations to a generative-AI tool; no specific tool named

Mata v. Avianca, Inc., No. 22-cv-1461 (PKC)

U.S. District Court for the Southern District of New York · June 22, 2023 · Verified against the court’s published document on July 19, 2026.

The foundational fabricated-citations sanctions decision (Castel, J.; reported at 678 F. Supp. 3d 443): Opinion and Order on Sanctions finding subjective bad faith by attorneys Peter LoDuca and Steven A. Schwartz — grounded in conscious avoidance and false and misleading statements to the court, not the ChatGPT use itself — and imposing a $5,000 penalty jointly and severally on both attorneys and Levidow, Levidow & Oberman P.C. The order dissects six fabricated decisions ("Varghese," "Shaboon," "Petersen," "Martinez," "Durden," "Miller"), including fake internal citations and a self-citing opinion, and requires respondents to notify their client and every judge whose name was falsely invoked. The court's much-quoted framing: "there is nothing inherently improper about using a reliable artificial intelligence tool for assistance," but existing rules "impose a gatekeeping role on attorneys to ensure the accuracy of their filings."

• Sanction: Monetary penalty payable to the Registry of the Court; mandatory notification letters to the client and to each judge falsely identified as the author of a fake opinion, with compliance filings
• Amount: $5,000, jointly and severally (two attorneys and their law firm)
• Authority: Fed. R. Civ. P. 11(b)(2), 11(c)(1), 11(c)(4), or alternatively the court's inherent power; a separate 28 U.S.C. § 1927 sanction was expressly declined
• AI tool alleged (per the primary document): ChatGPT, expressly — the court found the fake opinions were "created by the artificial intelligence tool ChatGPT," and respondents acknowledged the six fabricated decisions were ChatGPT-generated

What this dataset is, and is not

Every row is checked against the court’s own published order or opinion before it appears here — case name, docket, decision date, sanction amounts, and rule basis — and each row shows the date we last verified it. Candidates that have not yet passed that check stay out, no matter how widely they are reported. That makes this a verification-first dataset, not a census: for the broader landscape of fabricated-citation sanctions and the candor duties behind them, start with our guide to AI hallucination sanctions and Rule 3.3 rather than treating this list as complete.

The data is free to reuse under CC BY 4.0: cite “Legal AI Compliance Tracker (legalaicompliance.help)” and link back. The JSON and CSV files are regenerated on every site build, so they always match this page. Spot an error or a case we should verify next? Email info@mhsbsolutions.com; corrections ship within a week and are recorded on the changelog.