AI Hallucination Sanctions: A Tracker of Key Cases

The AI hallucination sanctions lawyers face now range from fines of $1,000 to $10,000, to fee awards above $30,000, to disqualification, bar referrals and, in one 2026 case, a default judgment against the client. Courts have not settled on a standard, and judges openly disagree about how harsh to be. What every ruling below shares is that the sanction followed a failure to check citations, not the use of AI itself.

How Many AI Hallucination Cases Have Courts Found?

The most widely cited count is Damien Charlotin’s AI Hallucination Cases database, which listed 2,207 cases when we read it on October 23, 2026. It tracks decisions where a court or tribunal found or implied that a party relied on hallucinated material, not every filing that used AI. Its filters attribute 1,267 entries to self-represented litigants and 877 to lawyers, with smaller numbers for judges (32) and experts (17).

The growth is steep. Writing in the ABA’s Judges’ Journal on November 3, 2025, retired Judge Herbert B. Dixon Jr. counted 426 cases in the same database as of early October 2025, starting with two in May 2023. His manual review found record monthly counts climbing from 2 in May 2023 to 73 in July 2025. By our reading, the total roughly quintupled in the following year.

How Large Are the AI Hallucination Sanctions Lawyers Face?

The early benchmark is Mata v. Avianca. On June 22, 2023, Judge P. Kevin Castel of the Southern District of New York imposed a $5,000 penalty, jointly and severally, on two lawyers and their firm for submitting fake ChatGPT-generated opinions and standing by them, and ordered letters to the client and to each judge falsely named as an author. The court wrote that there is “nothing inherently improper about using a reliable artificial intelligence tool for assistance.”

Fines stayed modest through early 2025. In Wadsworth v. Walmart, the District of Wyoming found that motions in limine cited nine cases, eight of which did not exist. Its February 24, 2025 order revoked the drafting lawyer’s pro hac vice admission, fined him $3,000, and fined two other lawyers $1,000 each, including local counsel.

The numbers then rose. In Noland v. Land of the Free, California’s Second District Court of Appeal found that 21 of the 23 case quotations in an opening brief were fabricated, and in its September 12, 2025 opinion ordered the lawyer to pay $10,000 and directed the clerk to send the opinion to the State Bar. In the Flycatcher trademark case in the Southern District of New York, Judge Katherine Polk Failla found that 14 of 60 cases in a brief did not exist, entered default judgment against the client on February 5, 2026, and on October 9, 2026 ordered the lawyer to personally pay $38,877.59 to the opposing lawyer who caught the fakes, as Greenberg Traurig summarized on October 20, 2026.

Are the Sanctions Proportionate?

Judges are split. In Johnson v. Dunn, Judge Anna M. Manasco of the Northern District of Alabama wrote in a July 23, 2025 order that the reprimands and modest fines that had become common were proving “insufficient deterrents.” She publicly reprimanded three lawyers who confirmed that ChatGPT had produced the citations, disqualified them from the case and referred the matter to the Alabama State Bar. The same order cites Lacey v. State Farm, a May 5, 2025 Central District of California decision ordering the plaintiff’s law firms to pay $31,100 in fees and costs.

Judge Dixon’s Judges’ Journal article describes the pushback. It summarizes an essay by retired Judge Ralph Artigliere and Professor William F. Hamilton arguing that the Johnson sanctions may have been disproportionate because the citations did not misstate the law and caused no substantive prejudice, and it reports that Judge Xavier Rodriguez has suggested some judges have overreacted. The article notes that other judges see unverified AI use as reckless and see no need for moderation. Artigliere and Hamilton propose weighing harm, culpability, remediation and deterrence.

The pattern across these rulings is consistent even where the severity is not. Courts escalate when a lawyer keeps filing fake citations after a warning, as in Flycatcher, or makes false statements about what happened, as in Mata. Supervising and local counsel were sanctioned alongside the drafter in both Wadsworth and Johnson.

What Should You Watch Next?

Watch whether appellate courts set a standard for proportionality, since trial judges are currently writing their own. Watch fee-shifting to opposing counsel, which Flycatcher showed can exceed any fine. And watch the database count itself; at the current pace, monthly totals keep setting new highs.

Common Questions

What was the first major AI hallucination sanction?

Mata v. Avianca, where the court imposed a $5,000 penalty on June 22, 2023.

How many AI hallucination cases are there?

Charlotin’s database listed 2,207 as of October 23, 2026, including 877 involving lawyers.

Can a lawyer be disqualified for AI citations?

Yes. In Johnson v. Dunn on July 23, 2025, three lawyers were disqualified from the case.

Do courts ban AI use in filings?

Not in these rulings. Mata said using a reliable AI tool is not inherently improper, and Flycatcher’s court said it does not oppose AI-assisted research.