Matthew A. Mishak, Esq.
Founder, LegalTek.ai • AI Governance & Legal Technology
Executive Summary
U.S. courts have moved from cautionary language about generative AI to actual monetary sanctions, attorney disqualification, pro hac vice revocations, and referrals to disciplinary authorities when lawyers file briefs containing hallucinated or fabricated citations that were not independently verified.
Across these orders, the judicial message is consistent: "if you sign it, you own it—and 'AI made it up' is not a defense to inadequate inquiry, lack of supervision, or lack of candor."
Courts have treated AI output as just another input that must be validated, much like the work product of a nonlawyer assistant or outsourced drafter.
Ethics Baseline for Lawyers Using Generative AI
ABA Formal Opinion 512
Frames generative AI as a tool that can improve efficiency but does not dilute baseline professional obligations—particularly competence, confidentiality and data exposure, supervision, candor to tribunals, client communications, and reasonable fees.
NYC Bar Formal Opinion 2024-5
Gives New York–specific guidance anchored in the NY Rules of Professional Conduct. Emphasizes "guardrails" rather than entirely new rules—because existing duties already govern the problem.
These ethics frameworks matter because many sanctions orders explicitly connect hallucinated citations to Rule 11 "reasonable inquiry" duties, inherent-authority sanctions, appellate sanctions authority, and professional conduct expectations.
Enforcement Actions: The Complete Record
Reported U.S. matters through February 11, 2026 in which courts imposed sanctions, fines, discipline referrals, or other concrete enforcement measures after filings contained hallucinated/fabricated authorities connected to generative AI.
Mata v. Avianca, Inc.
S.D.N.Y. (No. 1:22-cv-1461)
Lawyers filed a brief citing multiple nonexistent decisions generated via ChatGPT and then compounded the problem by filing follow-up submissions that still relied on fabricated authority instead of verifying.
Park v. Kim
2d Cir. No. 22-2057
Counsel cited a nonexistent precedent; the court highlighted the basic duty to verify citations before filing and referred the matter for professional discipline review.
Gauthier v. Goodyear Tire & Rubber Co.
E.D. Tex. No. 1:23-cv-00281
Plaintiff's counsel filed a response containing fabricated case citations and quotations produced by an AI research tool and failed to confirm accuracy before filing.
United States v. Hayes
E.D. Cal. No. 2:24-cr-0280-DJC
Defense counsel cited a hallucinated case and attributed quotations that did not exist; the court found counsel's follow-up explanations misleading and treated the conduct as sanctionable.
Nguyen v. Savage
E.D. Ark. No. 4:24-cv-00815-BSM
Plaintiff's filing cited four cases that did not exist; after a show-cause order, counsel pointed to artificial intelligence and workload/personal issues as contributing factors.
Wadsworth v. Walmart Inc.
D. Wyo. No. 2:23-cv-00118-KHR
Multiple motions cited numerous nonexistent cases; the court emphasized that even attorneys who permitted signature oversight still had a nondelegable duty to ensure filings are supported by existing law.
Mid Central Operating Engineers v. HoosierVac LLC
S.D. Ind. No. 2:24-cv-00326
Counsel submitted filings containing nonexistent authorities and failed to satisfy Rule 11's reasonable-inquiry duty; the order reflects courts' increasing willingness to sanction the attorney personally.
Dehghani v. Castro
D.N.M.
Counsel outsourced a brief via a freelance platform and did not verify cited cases or quotations; multiple cited authorities were nonexistent or did not support the propositions asserted.
Bunce v. Visual Technology Innovations
E.D. Pa. No. 2:23-cv-01740
Counsel submitted fabricated citations using ChatGPT, leading to a monetary sanction and CLE requirement. Frequently cited as a Third Circuit–region example of sanctionable AI hallucinations.
Johnson v. Dunn
N.D. Ala. No. 2:21-cv-01701-AMM
Defense counsel conceded the cited cases were ChatGPT hallucinations and that no verification occurred before filing. The court imposed escalating sanctions including removal from the case.
Safe Choice, LLC v. City of Cleveland
N.D. Ohio
After issuing a show-cause order, the court found Rule 11 violations and imposed sanctions; it also directed service of the sanctions materials to a local bar grievance committee.
Wilson v. KIPP
N.D. Tex. No. 3:24-cv-01638
The court sanctioned counsel after a declaration and briefing included artificially generated quotations presented as accurate. Illustrates that hallucinations are not limited to case citations—they include fabricated record quotations.
Lexos Media IP v. Overstock
D. Kan. No. 22-cv-2324
The court fined multiple attorneys after pleadings included fabricated and misrepresented legal authority submitted without adequate verification. Sanctions allocated based on each lawyer's role.
Amarsingh v. Frontier Airlines
10th Cir. No. 24-1391
The Tenth Circuit imposed sanctions on appeal after briefing relied on authorities that did not meet minimum standards of accuracy, citing the growing body of AI-hallucination sanction cases as context.
Enforcement Timeline
Jun 2023
Mata v. Avianca — First major AI sanctions
Jan 2024
Park v. Kim — First appellate discipline referral
Nov 2024
Gauthier v. Goodyear — Monetary sanction + mandatory CLE
Jan 2025
U.S. v. Hayes — Criminal case sanctions + bar notice
Feb 2025
Wadsworth v. Walmart — Pro hac vice revocation
May 2025
HoosierVac — $6,000 personal attorney sanction
Jul 2025
Johnson v. Dunn — Disqualification + bar referral
Oct 2025
Safe Choice v. Cleveland — Grievance committee notice
Feb 2026
Lexos v. Overstock — $12,000 multi-attorney fines
Feb 2026
Amarsingh v. Frontier — 10th Circuit appellate sanctions
Near-Miss and Advisory Orders
Not every hallucination incident ends in monetary sanctions. Courts have also issued show-cause orders, stern admonishments, and "warning opinions" meant to deter future misconduct—often emphasizing that the court may sanction more harshly next time.
NYC Bar Formal Opinion 2024-5 cites U.S. v. Cohen (S.D.N.Y. 2024) as an example where filings cited "cases that do not exist" after the client provided citations hallucinated by a public AI tool, and counsel failed to check them before submitting. Several later sanctions decisions treat earlier "no-sanction" outcomes as near-miss warnings.
Governance Analysis Through the COUNSEL Framework
Across the enforcement actions, the failure mode is rarely "AI exists." It is governance failure—in controls, supervision, and verification discipline—inside a litigation workflow. The repeated patterns map cleanly to COUNSEL:
Confidentiality
Not the typical trigger in hallucination sanctions orders, but ethics opinions make clear that confidentiality risk rises when client data is entered into third-party AI systems without adequate safeguards or consent.
Oversight
Central when lawyers delegate drafting, outsource work, or allow signature delegation without review. Courts have treated this as a 'nondelegable duty' problem—signers sanctioned even if they did not draft.
Understanding
Shows up whenever counsel claims they did not appreciate hallucination risk or did not know how a tool works. Both ethics opinions and sanctions orders reject 'I didn't know the tool makes things up' as incompatible with competence.
Notice
Appears in obligations to correct the record, respond candidly to OSCs, notify clients, and—when ordered—notify courts or disciplinary authorities. Several sanctions orders explicitly include service on clients and bar entities.
Scrutiny
The core: independent verification of citations, quotations, and factual assertions before filing. Nearly every sanctions order turns on the failure to perform this minimum verification step.
Equity
Seldom the explicit basis of these sanctions orders, but equity obligations remain relevant where AI use can embed bias in downstream advice or decision-making. The NYC Bar opinion explicitly includes nondiscrimination concerns.
Learning
Increasingly built into sanctions remedies: courts frequently order CLE or other education requirements, reflecting a judicial view that deterrence requires ongoing professional training.
Methodology and Limitations
This report covers developments through February 11, 2026. Primary-source collection prioritized official and semi-official repositories (including GovInfo, court-hosted PDFs, and Justia dockets/opinions) and then corroborated context with reputable legal news reporting.
Inclusion criteria: (i) U.S. federal or state court matter, (ii) attorney involvement, (iii) an order imposing sanctions/fines/referrals or issuing explicit judicial criticism tied to fabricated citations/quotes connected to generative AI, and (iv) a reasonably accessible primary document.
Limitations: There is no single authoritative "exhaustive" public database of AI-hallucination sanction orders; some relevant orders remain behind PACER or commercial reporters, and some state trial-level actions are difficult to retrieve.
Key Sources & References
- ABA Formal Opinion 512 – First AI Ethics Guidance
- NYC Bar Formal Opinion 2024-5 – Generative AI in the Practice of Law
- Mata v. Avianca, Inc. — S.D.N.Y. Sanctions Order
- Park v. Kim — 2d Circuit Discipline Referral
- Nguyen v. Savage — E.D. Ark. Rule 11 Order
- Safe Choice, LLC v. City of Cleveland — N.D. Ohio Sanctions
- Amarsingh v. Frontier Airlines — 10th Circuit Appellate Sanctions
Disclaimer: This content was human-reviewed but may contain AI-generated elements. Readers should conduct their own research and remain skeptical of factual errors. This article is provided for educational purposes and does not constitute legal advice.