How Many Lawyers Have Been Sanctioned for AI-Hallucinated Citations

Written by the Citation Safe Research Desk · Reviewed by Andy Gaber, Founder — July 16, 2026

The single most-searched question in this space deserves a number, not a vibe. The most reliable public source is the Damien Charlotin AI Hallucination Cases database, which as of its most recent public update tracks well over 1,700 documented instances of AI-fabricated or misused citations reaching a court, tribunal, or agency, across dozens of countries.

Why there is no single official count

No court system or bar association maintains a centralized, mandatory reporting requirement for AI-hallucination incidents, so every public count — including Charlotin's — is necessarily built from manually identified, publicly available orders and news coverage. That means the true number is almost certainly higher than any published tracker, since incidents caught informally by opposing counsel and corrected before a written order are rarely reflected.

What the trend line shows

The rate of newly documented cases has not meaningfully slowed since Mata v. Avianca made mainstream news in 2023. If anything, the pace of new entries has increased as AI drafting tools have become more widely adopted across the profession, including by pro se litigants who have no institutional research support at all.

Who gets sanctioned

The documented cases span solo practitioners, mid-size and large firms, in-house counsel, and pro se litigants. Notable examples include Mata v. Avianca (S.D.N.Y. 2023), Park v. Kim, 91 F.4th 610 (2d Cir. 2024), Wadsworth v. Walmart Inc. (D. Wyo. 2025), Coomer v. Lindell (D. Colo.), and Kruse v. Karlen (Mo. Ct. App. 2025), a pro se case that drew a $10,000 sanction. No category of filer appears to be immune.

Expert perspective

Charlotin has noted in commentary accompanying the database that the growth in documented cases outpaces any plausible growth in AI tool adoption alone, suggesting that detection and reporting have also improved as courts and opposing counsel have gotten better at spotting the pattern.

What this means for your own practice

If you assumed this was a rare, one-off problem confined to a few high-profile cases, the data says otherwise. Treat citation verification as a mandatory step on every AI-assisted filing, not a precaution reserved for unusually large or high-stakes matters.

A common question

Is this only a U.S. problem?

No. The Charlotin database documents cases from multiple countries, including the UK, Canada, and Australia. See our overview of international AI hallucination sanctions cases for details outside the U.S.

Why the true number is almost certainly higher

Every publicly tracked count, including the Charlotin database, is built from written orders, published opinions, and news coverage that reached public visibility. A large and fundamentally unknowable share of AI-hallucination incidents are caught informally — opposing counsel flags an unlocatable citation in a phone call, the attorney quietly withdraws and refiles the brief, and no written order is ever issued. These informal corrections never enter any public tracker, which means the documented case count should be read as a floor, not a ceiling, on how often this actually happens.

Breaking down the numbers by outcome severity

Of the documented cases, the outcomes span a wide range: some result only in a stricken filing with no further consequence, some add monetary sanctions, and a smaller subset add bar referrals or disciplinary proceedings. The distribution skews toward the less severe end for first-time, promptly-corrected incidents, and toward the more severe end for repeated incidents or cases where the attorney defended the fabricated citations after being challenged rather than correcting them.

What the growth rate implies for the next few years

Given that AI drafting tool adoption across the legal profession is still increasing, and that detection methods (both by opposing counsel and by court staff) have gotten more attentive rather than less, there is no structural reason to expect the count of documented cases to plateau in the near term. Building a verification habit now, before an attorney's own name potentially becomes an entry in a tracker like this, is meaningfully cheaper than the alternative.

Comparing this to other well-known error categories in litigation

For context, missed filing deadlines and conflicts-of-interest violations have long been the leading categories of malpractice claims industry-wide, tracked carefully by carriers for decades. AI-hallucination incidents are a comparatively new category, but the growth rate of documented cases since 2023 suggests it could become a similarly significant loss category within a few years if verification practices do not become as standardized as calendaring and conflicts-checking already are.

Why firm size doesn't predict immunity

It would be reasonable to assume large firms, with more resources and more layers of review, would be underrepresented in the documented cases. The record does not support that assumption. Large-firm attorneys appear in the Charlotin database alongside solo practitioners, suggesting that the number of people who could have caught an error matters less than whether any one of them was actually assigned to run an independent verification check before filing.

Final takeaway

The exact count of sanctioned attorneys matters less than the trend: this is not shrinking, and no category of practitioner has proven immune. A five-minute verification habit, run consistently, is the entire difference between being a footnote in next year's numbers and simply avoiding the problem altogether. Build the habit before a deadline forces the decision for you.

If your practice handles a high volume of similar filings, consider maintaining a shared, internally verified library of frequently-cited authority, updated whenever a new citation is added, so the verification cost is paid once per citation rather than once per filing. This single change has measurably reduced verification friction in high-volume practices without reducing rigor.

Bookmark the Charlotin database itself and check it periodically; it is updated as new cases are documented and is the single best free source for staying current on this issue as it develops.