AI sanctions tracker

Courts are sanctioning lawyers for AI-fabricated citations. Here is the record.

1,782

Court cases involving AI-hallucinated citations or content documented in the public AI Hallucination Cases database as of July 18, 2026 — up from roughly 200 a year earlier, now growing about 8 per day. The largest penalty to date: roughly $110,000.

source: AI Hallucination Cases database (damiencharlotin.com) · growth reporting: ABA Journal

last reviewed by Lavir: July 18, 2026 · figures change as the database grows — follow the source link for today's count

Notable cases

How it actually goes for the lawyer.

A non-exhaustive selection of documented decisions, described conservatively. Full dockets and hundreds more incidents are catalogued in the public database.

D. Or. sanctions opinion (Dec. 2025) — the record penalty

Fifteen nonexistent cases and eight fabricated quotations across three briefs; the court called the conduct "a notorious outlier in both degree and volume."

outcome: ≈$110,000 in fees, fines, and costs — the largest AI-citation penalty reported to date

Lacey v. State Farm (C.D. Cal. 2025)

A brief with roughly a third of its citations wrong or nonexistent — prepared using Westlaw's own AI tools (CoCounsel, Westlaw Precision) and filed by two major firms, including the 14th-largest in the country, without an independent check.

outcome: ≈$31,000 in sanctions, jointly — proof that owning the tools is not the same as verifying

Mata v. Avianca, Inc., 678 F. Supp. 3d 443 (S.D.N.Y. 2023)

The case that started it: a brief citing multiple nonexistent decisions generated by ChatGPT, defended after opposing counsel couldn't find them.

outcome: $5,000 sanction · national coverage that followed the attorneys' names

Park v. Kim, 91 F.4th 610 (2d Cir. 2024)

A reply brief cited a fake case produced by ChatGPT; the court of appeals addressed counsel's AI use directly.

outcome: referral to the circuit's attorney grievance panel

Gauthier v. Goodyear Tire & Rubber Co. (E.D. Tex. 2024)

Response briefing included nonexistent cases and fabricated quotations attributed to real ones.

outcome: monetary sanction · mandatory continuing legal education on generative AI

Wadsworth v. Walmart Inc. (D. Wyo. 2025)

Motions in limine cited cases that did not exist, produced by a firm's internal AI tool — showing the risk isn't limited to public chatbots.

outcome: monetary sanctions · one attorney's pro hac vice admission revoked

Johnson v. Dunn (N.D. Ala. 2025)

Filings by large-firm defense counsel contained AI-fabricated citations.

outcome: public reprimand · disqualification from the case · bar referrals

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

A published appellate opinion found the overwhelming majority of quotations in the briefs were fabricated — and warned the bar directly.

outcome: $10,000 sanction · published opinion naming the conduct

Case descriptions summarize public decisions and coverage; consult the linked database and the underlying dockets for full detail. Professional-conduct guidance: New York State Bar Association.

The pattern

It's never the AI that gets sanctioned.

Every consequence above landed on the human who signed the filing: fines, fee awards, mandatory education, revoked admissions, bar referrals, published opinions carrying the lawyer's name. Courts have been consistent on the principle — using AI is not the violation; filing unverified AI output is.

The fix is procedural, not technological: verify every citation, quotation, and holding against primary sources before filing — and keep a record showing you did.

Stay off this page

Verify before you file.

Lavir independently checks every citation in your filing against public court records and corrects the outliers before a judge finds them. Your first filing is free.

Check one filing free see a sample report