ai fabricated citations in 2026: what the court and tribunal sanctions record shows
2026-08-06
Lawyers keep filing briefs with AI-invented cases, and courts keep sanctioning them, from $3,000 to $31,150 and up. Four documented rulings across two countries show the pattern and the one thing every panel now agrees on: the AI excuse makes it worse, not better.
Built on verified case files. The argument below leans on evidence The Internet Ninja validated against the public record and published in full, method included.
- Ontario Law Society Tribunal orders a lawyer to pay $31,150 in costs after Grok-generated fake citations
- Whiting v. City of Athens: a federal appeals court fines two lawyers $15,000 each for briefs with 'over two dozen fake citations'
- Torres Campos v. Munoz: a published California appellate opinion sanctions a lawyer $5,000 for citing AI-hallucinated cases in a dog-custody fight (2026)
- OTG New York v. Ottogi America: a New Jersey federal court fines a lawyer $3,000 for a reply brief built on AI-hallucinated citations (2025)
Four sanctions, two countries, amounts from $3,000 to $31,150. The pattern across all of them: no court or tribunal on this record has treated “the AI made it up” as an excuse. Several have treated it as an aggravating factor.
What “ai fabricated citations” means in practice
An AI-fabricated citation is a case reference, invented outright or real but misdescribed, that a generative AI tool produced and a lawyer filed without independently checking it against a primary source. The failure isn’t that AI hallucinates; every current model does, sometimes. The failure sanctioned in every one of these cases is filing without checking.
Four rulings, four amounts
Ontario, 2026. Shahryar Mazaheri filed tribunal motions built on citations to “non-existent decisions, or decisions that exist but do not stand for the propositions outlined.” He admitted the citations came from Grok. The Law Society Tribunal’s Hearing Division called the AI misuse “an additional and significantly aggravating factor” and ordered $31,150 in costs, with no reduction (2026 ONLSTH 112). TIN’s case file on Mazaheri has the full ruling detail.
Sixth Circuit, 2026. A federal appeals court found briefs containing “over two dozen fake citations and misrepresentations of fact,” by the court’s own conservative estimate, and fined two attorneys $15,000 each, $30,000 combined, plus full appellate fees, double costs, and a disciplinary referral. TIN’s case file on Whiting v. City of Athens documents the sanction.
California, 2026. In a published opinion, an appellate court sanctioned counsel $5,000 under Cal. Rules of Court rule 8.276(a)(4) after she conceded at oral argument that fabricated citation details may have come from AI tools. The court certified the opinion for publication specifically to warn against AI-hallucinated citations. TIN’s case file on Torres Campos v. Munoz has the ruling.
New Jersey, 2025. A federal court fined an attorney $3,000 under Rule 11(c) after fabrications in a reply brief were found to be “derived from generative artificial intelligence.” The brief was struck, and the attorney was ordered to self-report to his state bar. TIN’s case file on OTG v. Ottogi has the full order.
| Case | Jurisdiction | Sanction | AI tool named |
|---|---|---|---|
| Mazaheri v. Law Society of Ontario | Ontario tribunal | $31,150 costs, no reduction | Grok |
| Whiting v. City of Athens | 6th Cir. | $15,000 each (2 attorneys) | not named |
| Torres Campos v. Munoz | Cal. Ct. App. | $5,000 | not named |
| OTG New York v. Ottogi America | D.N.J. | $3,000 | not named |
What the pattern actually shows
Sanction size tracks volume and severity of fabrication, not remorse. Mazaheri’s tribunal explicitly named the AI use itself as aggravating. None of the four rulings reduced a sanction for candor about AI use; several increased scrutiny once AI use was confirmed. The one variable that consistently helped defendants was independent verification before filing, which is exactly what none of these four briefs had.
The bottom line
If your workflow includes generative AI for legal research, the sanctions record is now specific and growing: verify every citation against a primary source before it’s filed, because no court on this record has accepted “the AI did it” as mitigation, and at least one has called it aggravation.
Sources
- CanLII, 2026 ONLSTH 112, Law Society Tribunal, Hearing Division. https://www.canlii.org/en/on/onlst/doc/2026/2026onlsth112/2026onlsth112.html
- TIN case files: Mazaheri v. Law Society of Ontario, Whiting v. City of Athens, Torres Campos v. Munoz, OTG New York v. Ottogi America (full source lists on each page).
Questions
Why do lawyers keep filing AI-fabricated citations?
Generative AI tools produce fluent, correctly formatted case citations that don't exist or don't say what the brief claims. Lawyers who don't independently verify every citation against a primary source, CanLII, Westlaw, PACER, before filing are the ones getting sanctioned; the tool doesn't check its own work.
Does it matter which AI tool was used?
Documented sanctions name several: Grok in the Ontario Mazaheri case, unnamed generative AI tools in the Sixth Circuit's Whiting case and the New Jersey Ottogi case, and AI tools generally in the California Torres Campos case. The tool varies; the sanction pattern, verify before filing or answer for it, does not.
Does admitting AI use reduce the sanction?
No documented case in this record shows a reduction for admitting AI use. The Ontario tribunal in Mazaheri explicitly called the AI misuse 'an additional and significantly aggravating factor' and awarded full costs with no reduction.
Sources
- CanLII, 2026 ONLSTH 112, Law Society Tribunal, Hearing Division , 2026
This is analysis, not a verified outcome. It carries no verification badge and never will. The proof lives in the case files, where every figure is checked against the public record and the method is printed on the page.