Skip to main content
Citation Safe
← All posts

Couvrette v. Wisnovsky: The $110,000 Oregon AI Hallucination Sanction

Written by the Citation Safe Research Desk · Reviewed by Andy Gaber, Founder — July 19, 2026

If you want a single case that answers “how bad can this actually get?” when a lawyer files AI-fabricated citations, it is Couvrette v. Wisnovsky, decided in the U.S. District Court for the District of Oregon on March 23, 2026. The sanctions total, on the public record, is $110,204: a $15,504 monetary sanction against counsel, plus a $94,700 adverse-costs order shifting the opposing side’s fees. The court also struck the offending briefs and dismissed the underlying claims with prejudice. As of the July 2026 update of Damien Charlotin’s public AI Hallucination Cases database (~1,750 cases tracked), this is the largest single dollar exposure a U.S. lawyer has publicly suffered for filing AI-generated fake citations.

Why this case matters even if you never practice in Oregon

Couvrette is the case cite you hand a partner, a client, or a malpractice underwriter when they ask why AI-verification controls are worth paying for. Every prior sanction in this line — Mata v. Avianca in S.D.N.Y. ($5,000 combined, 2023), Kruse v. Karlen in the Missouri Court of Appeals ($10,000, 2024), Wadsworth v. Walmart in D. Wyoming (2025) — was small enough that a firm could rationalize it as a headline risk, not a P&L event. Couvrette blows past that threshold. Six figures out-of-pocket, on top of claim dismissal, changes the risk math for the smallest solo practice as well as for the largest firm.

The three sanctions layers stacked in Couvrette

The Oregon order layered three separate remedies, and that layering is what pushed the exposure past $100,000. First, a Rule 11-style direct monetary sanction against counsel — the sort of “pay the court” number that appears in most sanctions cases in this category, typically $1,000-$10,000. Second, an adverse-costs order, shifting opposing counsel’s reasonable attorney fees to the sanctioned side. This is the number that scales with the case’s size and posture: in Couvrette it was $94,700, which is not extraordinary for a defended civil action but is extraordinary when it flows from an AI-citation problem rather than the merits. Third, striking the briefs and dismissing the underlying claims with prejudice — the substantive loss, which nobody prices in when they estimate “the sanction.” The client’s case is gone. That is the fourth and often largest cost, and it is not a line item in any of the reported numbers.

Adverse-costs orders are the number to watch

If you scan Charlotin’s database looking for outlier penalties, the pattern that repeats is not enormous direct fines — most direct fines cluster $500 to $5,000. The pattern is that adverse-costs orders scale with the litigation’s size and can dwarf the fine. LiveVideo.AI Corp. v. Redstone (S.D.N.Y., June 2026) hit $80,056 that way. Joel A. Rivera v. Triad Properties Corporation (N.D. Alabama, March 2026) hit $35,603. Whiting v. City of Athens (6th Cir., March 2026) hit $30,000. Every one of these is fee-shifting, not a flat fine. A firm that treats “AI sanctions” as “a few thousand dollars, worst case” is planning around the wrong number.

Claim dismissal with prejudice is the quiet part

A prejudice dismissal ends the client’s right to refile the same claim. In Couvrette, whatever the merit of the underlying dispute, the client is done. Malpractice exposure from that outcome is downstream of the sanctions order and is not on the sanctions balance sheet. It is on your malpractice carrier’s balance sheet. Ask any malpractice underwriter what they think of AI-generated brief risk in the 2026 renewal cycle. The answer is now specific, dollar-quantified, and unfavorable.

The prevention math is trivial next to $110,204

A citation-verification tool, in-house or vendor-provided, that costs even a few hundred dollars per month per attorney runs $2,000-$10,000/year all-in for most solo-through-midsize firms. Set against Couvrette’s $110,204 out-of-pocket exposure and the downstream claim-dismissal and malpractice consequences, the prevention math is not close. What the market lacked before Couvrette was one clean case cite that made the number impossible to hand-wave. That case now exists on the public docket.

What every firm should do this week

Three things: (1) put a written verification requirement in your AI-use policy that specifically covers every citation and every quoted passage before filing, with a name attached to sign-off; (2) pick a verification method — either a documented manual workflow through CourtListener or a purpose-built tool — and make it non-optional, not aspirational; (3) log completion of verification in your matter file for every filed brief, so that if a Couvrette-shaped question ever comes at you, you have contemporaneous evidence of the check.

Common questions

Is $110,204 a one-off, or should we expect this magnitude to become routine?

The direct-fine component is likely to stay in the low thousands across most cases. The adverse-costs component scales with the litigation posture. Expect to see more six-figure combined numbers, especially in defended commercial cases where opposing counsel’s recoverable fees are large.

Would a citation-verification tool have caught the specific issue in Couvrette?

The public order does not walk through each fabricated citation, so we cannot say category-by-category. What is knowable: three-layer deterministic checkers (existence, quotation-match, proposition-support) catch the majority of the failure modes documented across the Charlotin database — invented citations, real cases with fabricated quotations, and real cases cited for a proposition they do not support.

Verify a brief before you file it →

Analysis & Learnings

Working forward from the docket facts — $110,204 combined ($15,504 direct + $94,700 adverse costs), briefs struck, claims dismissed with prejudice, fifteen fabricated citations plus seven misquotes across ECF Nos. 142, 155, and 168 — a bar-hearing-grade postmortem for Couvrette v. Wisnovsky looks like this.

The verification step that would have caught this

Every one of the fifteen fabricated citations is caught by a literal existence check against a public reporter index — a step measured in seconds per cite, not minutes. The seven misquotes are caught by a literal-string search of the quoted passage within the cited opinion’s text; the search returns no hit when the quotation was manufactured. A three-check verifier (existence, quotation-match, proposition-support) run before filing eliminates the entire fact pattern that produced the sanction.

Pattern this fits across the 2026 docket

Couvrette sits at the top of the 2026 ‘adverse-costs-dominant’ pattern: the $94,700 adverse-costs component is roughly six times the direct fine. The same shape shows up in LiveVideo.AI Corp. v. Redstone (S.D.N.Y., June 2026, $80,056 almost entirely adverse costs) and in the smaller Joel A. Rivera v. Triad Properties Corp. (N.D. Alabama, March 2026, $35,603 adverse-costs-dominant). Firms budgeting for ‘a few thousand dollars, worst case’ are planning around the wrong line item.

Concrete process changes a bar hearing report would want to see

  • A written pre-filing verification requirement in the firm AI-use policy that names a specific attorney as signer for every filed brief.
  • A contemporaneous verification log stored in the matter file, showing which tool ran, when, and against which reporter set — so the file survives a show-cause order intact.
  • Second-attorney sign-off on any adversarial civil filing over a set page count, with the second attorney separately certifying that every quoted passage was string-matched to the source.

Cost math against a $29/mo verification subscription

At $29/month Solo, Citation Safe costs $348 per year per attorney. Couvrette’s $110,204 out-of-pocket exposure is 316 years of Solo subscription — 92 years even at the $99/mo Suite tier — before you count the malpractice premium impact of a with-prejudice dismissal or the underlying claim value the client will never recover.

What NOT to do in the aftermath

Do not respond by banning AI firmwide. Bans move usage into shadow-tools attorneys still touch on their phones, and remove the paper trail your malpractice carrier will ask for. Ban unverified filings, not the drafting tool — the enforceable rule is that nothing gets filed that has not passed a documented citation check.


About Citation Safe

Citation Safe verifies every case citation and every quoted passage in a legal brief against public court records before you file. Three deterministic checks run in about ninety seconds: existence (does the case exist), quotation-match (does the exact quoted string appear in the reporter), and proposition-support (does the cited passage support the argument). Pricing is $29/mo Solo and $99/mo Suite. Start a free brief check →

Verify your own brief in 30 seconds

Every citation checked against primary sources — free, no card required.

Try free →

Get the weekly Legal AI Accuracy Report

Our live False-Verify Rate, dispute counts, and the Legal AI Accuracy Index — in your inbox weekly.

See the live scorecard →
Get the free Legal AI Vendor Comparison Report

Which legal AI tools publish their citation error rate — and which don't. Delivered by email, no spam.