AI sanctions and court rules

Courts sanction lawyers for filing unverified AI output and for a lack of candor once the errors surface, not for using AI. Penalties have grown from a $5,000 fine in 2023 to about $110,000 and a dismissal with prejudice in 2026, and now include removal from cases, bar referrals and suspensions.

Why do courts sanction lawyers for AI errors?

Because the filing contained fabricated citations or quotations that the lawyer did not check. ABA Formal Opinion 512 says such filings can breach Rules 3.1, 3.3 and 8.4(c), and that "Even an unintentional misstatement to a court can involve a misrepresentation" (p. 10).

The cases show a pattern (our analysis): lawyers who admitted the problem quickly and fixed it, as in Wadsworth and the firm in Johnson v. Dunn, were treated more lightly than those who stood by fake authorities.

What AI sanctions cases should lawyers know about?

The cases below are in date order. For each one we say whether we read the court's own document or relied on secondary reporting.

Mata v. Avianca, Inc. (S.D.N.Y., June 22, 2023)

Mata v. Avianca, Inc., No. 1:22-cv-01461-PKC, 678 F. Supp. 3d 443. Lawyers filed an opposition citing nonexistent opinions generated by ChatGPT, then stood by them after Judge P. Kevin Castel questioned them. The court found subjective bad faith.

Sanction: $5,000, jointly and severally, against Peter LoDuca, Steven A. Schwartz and Levidow, Levidow & Oberman, paid into the court registry within 14 days. They also had to send letters to their client and to each real judge falsely named as the author of a fake opinion.

Source: primary, the sanctions order (also on Justia).

Park v. Kim (2d Cir., Jan. 30, 2024)

Park v. Kim, No. 22-2057, 91 F.4th 610. The lawyer cited a nonexistent decision she got from ChatGPT and had not read.

Sanction: referral to the court's Grievance Panel, and an order to give her client a copy of the decision (translated into Korean if needed). No fine.

Source: the opinion on govinfo. Our summary has not been re-checked against the full text.

Wadsworth v. Walmart Inc. (D. Wyo., Feb. 24, 2025)

Wadsworth v. Walmart Inc., No. 2:23-cv-118. Motions in limine cited nine cases, eight of them fake. They came from Morgan & Morgan's in-house AI platform.

Sanction: the drafting lawyer's pro hac vice admission was revoked and he was fined $3,000. Two signing lawyers were fined $1,000 each.

Source: secondary reporting only (Today's General Counsel). We did not read the court's order.

Lacey v. State Farm General Insurance Co. (C.D. Cal., May 6, 2025)

Lacey v. State Farm General Ins. Co., No. 2:24-cv-05205-FMO-MAA. A brief from Ellis George and K&L Gates had about 9 bad citations out of 27, including two nonexistent authorities. A corrected brief still had errors, and the AI use was not disclosed. Special Master Michael R. Wilner (ret.) issued the order.

Sanction: $31,100, jointly and severally ($26,100 for Special Master fees plus $5,000 toward State Farm's fees). The briefs were struck.

Source: secondary reporting. An order PDF is available but we did not read it.

Johnson v. Dunn (N.D. Ala., July 23, 2025)

Johnson v. Dunn, No. 2:21-cv-1701-AMM, 792 F. Supp. 3d 1241. Butler Snow lawyers filed motions with ChatGPT-fabricated citations, even though the firm had AI policies.

Sanction: no fine. Judge Anna M. Manasco wrote that the court "REPRIMANDS ... DISQUALIFIES them from further participation in this case; and REFERS this matter to the Alabama State Bar and other applicable licensing authorities." Three lawyers were affected. The firm itself was not sanctioned.

Source: primary, the opinion (Berkeley Law copy).

Noland v. Land of the Free, L.P. (Cal. Ct. App., Sept. 12, 2025)

Noland v. Land of the Free, L.P., No. B331918, 114 Cal.App.5th 426. Of 23 case quotations in the appellate briefs, 21 were fabricated by AI tools. It is California's first published opinion on AI hallucinations.

Sanction: $10,000. Counsel had to serve the opinion on the client, and the clerk notified the State Bar.

Source: secondary reporting (McGuireWoods). The opinion on Justia was not re-read.

Couvrette v. Wisnovsky (D. Or., Dec. 12, 2025 and Mar. 23, 2026)

Couvrette v. Wisnovsky, No. 1:21-cv-00157-CL. Three briefs contained 15 nonexistent cases and 8 fabricated quotations. Magistrate Judge Mark D. Clarke issued an opinion on December 12, 2025 and a fee order on March 23, 2026.

Sanction: about $110,000 in total (fines plus opposing fees), and the case was dismissed with prejudice. The ABA Journal (Apr. 17, 2026) reports that Judge Clarke wrote: "If there was ever an 'appropriate case' to grant terminating sanctions for the misuse of artificial intelligence, this is it."

Source: secondary reporting only (ABA Journal).

Fletcher v. Experian Information Solutions, Inc. (5th Cir., Feb. 18, 2026)

Fletcher v. Experian Information Solutions, Inc., No. 25-20086. The court found that counsel "used artificial intelligence to draft a substantial portion, if not all, of her reply brief and then failed to verify the accuracy" and was "not forthcoming."

Sanction: $2,500 against attorney Heather Hersh.

Source: primary, the published opinion.

Whiting v. City of Athens, Tennessee (6th Cir., Mar. 13, 2026)

Whiting v. City of Athens, Tenn., Nos. 24-5918/5919/25-5424. The briefs contained more than two dozen fabricated citations and misstatements of the record.

Sanction: full appellate attorney fees and double costs, a $15,000 punitive sanction for each of the two lawyers, and referral for discipline. According to LawNext and Norton Rose Fulbright, the court did not expressly find that AI was used.

Source: secondary reporting only (LawNext, Norton Rose Fulbright).

Lnu v. Blanche (9th Cir., June 3, 2026)

Lnu v. Blanche, No. 24-4790. Attorneys Mike Singh Sethi and William Rounds filed briefs with "multiple nonexistent cases, misattributed quotations, and gross misrepresentations of real cases," and claimed the errors were typographical. The court said it was not sanctioning them for using generative AI: "the rules are not violated at the point of research and drafting, but at the point of signing and filing." It tied the severity of the discipline to their "repeated failure of candor."

Sanction: $2,500 for each lawyer; a six-month suspension from practice before the Ninth Circuit; a copy of the order to their clients, opposing counsel and the presiding judge in every pending case, and to every lawyer in their firm. For two years, the two lawyers and all attorneys at their firm must include in every filing a statement under penalty of perjury "addressing whether generative AI was used, disclosing the name of the tool used," and certifying that the signer personally reviewed the filing and that all citations and quotations refer to existing authority. The clerk was ordered to serve the order on the State Bar of California and any other applicable licensing authorities.

Source: primary, the order.

Do judges require lawyers to disclose AI use?

Some do. Individual judges' standing orders and some local rules require either a disclosure that AI was used in preparing a filing, or a certification that a human checked every citation, or both. Some cover all "AI", others only "generative AI".

ABA Formal Opinion 512 tells lawyers to check local court rules on AI disclosure before filing (p. 10, n. 47).

Judge Michael M. Baylson, E.D. Pa. (June 6, 2023)

His standing order covers all AI, not just generative AI. A party that "has used Artificial Intelligence ('AI') in the preparation of any complaint, answer, motion, brief, or other paper ... MUST, in a clear and plain factual statement, disclose that AI has been used in any way in the preparation of the filing, and CERTIFY, that each and every citation to the law or the record in the paper, has been verified as accurate."

Judge Brantley Starr, N.D. Tex. (May 30, 2023, since superseded)

His order was the first of its kind. It required a certificate that no generative AI was used, or that a human checked it. His chambers page now says he "no longer requires a certification on the use of artificial intelligence in light of the Court's adoption of Local Civil Rule 7.2(f) and Local Criminal Rule 47.2(e), requiring the disclosure of the use of generative AI." Read those local rules for the current requirement.

Other early orders

Judge Stephen Vaden (U.S. Court of International Trade) and Magistrate Judge Gabriel Fuentes (N.D. Ill.) also issued early AI orders that drew national attention (Gentry Locke). Check the current text with each court.

The Fifth Circuit chose not to adopt an AI rule

The Fifth Circuit proposed an AI-certification rule (5th Cir. R. 32.3) in January 2024, received 13 comments, and declined to adopt it because existing rules were sufficient (Fletcher opinion, p. 3).

How do I find out whether my judge has an AI order?

Read the assigned judge's individual practices and the court's local rules before every filing. Orders change, as Judge Starr's did.

Two trackers collect them: Bloomberg Law's comparison table (the free version covers orders through January 30, 2024; the live tracker requires a subscription) and Trace.law's Court AI Disclosure Requirements tracker. Treat them as starting points, not as the rule itself.

Is there a federal rule on AI evidence?

Not yet. Proposed Federal Rule of Evidence 707 would apply Rule 702's reliability standards when "the product of artificial intelligence is offered as evidence without an accompanying expert."

On May 7, 2026 the Advisory Committee on Evidence Rules "voted to withdraw the proposed rule as issued for public comment, and to submit a modified rule proposal for further study" (agenda book, Oct. 15, 2026, p. 5). Any new version would need another round of public comment.

Are "AI sanctions" really something new?

Mostly not (our analysis). The Sixth Circuit sanctioned fake citations in Whiting without finding that AI was used, and the Fifth Circuit declined an AI-specific rule because existing rules already cover the problem. The duty is the same one lawyers have always had: verify everything you file.


Last reviewed November 9, 2026. This is general information, not legal advice.

About Faradex. Faradex is a private AI workspace for small and mid-sized firms that handle confidential client work. Each firm gets its own Workspace on dedicated infrastructure in the United States. Files and conversations are held in memory and are gone when the Workspace is reset. AI model: Currently Opus via AWS Bedrock, more information is available on our website. AWS does not store prompts or responses, and model providers have no access to them. Pricing: from $120 a month, no seat minimum; see faradex.ai/pricing.