Oregon AI Sanction: $110K Against 2 Lawyers for Fake Cases

In Couvrette v. Wisnovsky, No. 1:21-cv-00157-CL (D. Or.), Magistrate Judge Mark D. Clarke ordered two lawyers to pay a combined $110,204.38: a $15,500 fine plus $94,704.38 of the defendants' attorney fees. The money followed summary-judgment briefs that cited 15 nonexistent cases and carried fabricated quotations. The judge also struck the briefs and dismissed the plaintiffs' claims with prejudice. The figures come from two orders, dated December 12, 2025 and March 23, 2026.

Reviewed and corrected on 12 October 2026. This is general information, not legal advice; check the cited orders before relying on any of it. The case is still moving, so see the status section below.

TL;DR

• Who owes what: Stephen Brigandi, the California lawyer admitted pro hac vice, owes $95,998.72: the $15,500 fine plus $80,498.72 in fees (85% of the fee award). Timothy Murphy, the Oregon local counsel, owes $14,205.66, the other 15%.
• What was filed: three summary-judgment briefs with 15 nonexistent cases and fabricated quotations falsely attributed to eight real authorities. The $15,500 fine prices 15 cases at $500 each and 8 quotations at $1,000 each.
• The rules: Rule 11(b) and District of Oregon Local Rule 83-7 for Brigandi; Local Rule 83-3, which requires local counsel to "meaningfully participate," for Murphy.
• Why it grew: most of the total is fee-shifting, and the court added a terminating sanction. It found that deleting the fake authority from the amended briefs did not fix the problem: "Rather than a correction, Mr. Brigandi attempted a coverup."
• Where it stands: the plaintiffs appealed to the Ninth Circuit on July 22, 2026 (No. 26-4712). The notice of appeal names the March 30, 2026 summary judgment opinion and the June 22, 2026 order denying reconsideration; we did not read the appeal briefs. As of June 12, 2026 Brigandi had not paid the $15,500 and was ordered to show cause why he should not be held in contempt.
• Is it the largest? In the Charlotin AI hallucination database (page last updated 10 October 2026), it is the largest monetary penalty among US entries. The next is ByoPlanet v. Johansson (S.D. Fla., 2025) at $85,567.75 in fee awards. We have not checked every source.

Part of our legal AI verification and hallucination guide series.

For related verification and hallucination coverage, see AI Hallucinations in Legal Research: How to Avoid Sanctions, Damien Charlotin's Hallucination Tracker, Read Like a Risk Manager and the tracker of sanction cases.

Primary sources, as read on 12 October 2026

• Citation: Couvrette v. Wisnovsky, No. 1:21-cv-00157-CL (D. Or.), Opinion and Order of December 12, 2025 (ECF 215, cited by the court as 2025 WL 4109655), and Opinion and Order of March 23, 2026 (ECF 225).
• Links: the December 12, 2025 order, the March 23, 2026 order and the docket.
• The court's own summary: "In the quickly expanding universe of cases involving sanctions for the misuse of artificial intelligence, this case is a notorious outlier in both degree and volume."
• What changed in 2026: the fee award was approved at $94,704.38 and split between the two lawyers on March 23, 2026, and the case went on to a summary judgment ruling, a contempt show-cause order and an appeal (all below).

The case in one paragraph

Couvrette started as a family dispute over the tenancy and ownership of a winery. Joanne Couvrette sued two of her brothers in January 2021 as trustee of her mother's estate. Stephen Brigandi, a lawyer in California (his filings give a San Diego address), appeared pro hac vice on February 7, 2024, with Timothy Murphy as Oregon local counsel. The court recorded that Brigandi's son was dating Couvrette's daughter and that Brigandi had agreed to represent the plaintiffs for free.

The plaintiffs' claims never got a merits ruling. They were dismissed as a sanction, after the defense pointed out fake cases in its own summary-judgment papers.

How the case unfolded

The December 12 order recounts the record in detail.

• January 31, 2025: the plaintiffs' summary-judgment motion cited two nonexistent cases.
• April 4, 2025: their response cited seven more. The defendants' brief the same day identified the first two and suggested they were "perhaps hallucinated by artificial intelligence."
• April 25, 2025: the plaintiffs filed a notice of errata. The court found that it did not acknowledge, explain or amend the passages that relied on nonexistent cases.
• May 2, 2025: the reply cited six more nonexistent cases (15 in total across three briefs) and contained fabricated quotations attributed to seven real cases and to the Restatement (Second) of Contracts.
• May 7 and 21, 2025: the defendants served a proposed Rule 11 motion on May 7. On May 21 Brigandi answered by letter that the plaintiffs "will not agree to the sanctions requested" and would respond with a notice of errata if a motion was filed. The defendants filed their motion on May 23.
• May 26, 2025: the plaintiffs filed amended briefs. The court found they deleted the nonexistent cases and fabricated quotations but left the statements of law in place without citation. In one instance only the quotation marks around a fabricated Restatement statement were removed.

The court's conclusion on that amended briefing: "Rather than a correction, Mr. Brigandi attempted a coverup. He failed at both." It also found no evidence that Brigandi or the plaintiffs took any step to verify the amended briefing.

Brigandi was hospitalized from June 30 to July 29, 2025. The court said it did not question his treating physician's credibility. It did not give full weight to the doctor's declaration, which was filed late, and it said the declaration did not explain his later conduct, including a motion to withdraw that he described as unopposed when the defendants opposed it. The court held a sanctions hearing on October 29, 2025.

What the court ordered on December 12, 2025

Under Rule 11 and the local rules the court:

1. Struck the plaintiffs' summary-judgment briefs and amended filings without leave to refile.
2. Ordered Brigandi to pay the Clerk $15,500.
3. Held that the defendants could recover reasonable fees and expenses caused by the violations, with the amount and split to be set after Murphy responded.
4. Vacated its earlier order letting Murphy withdraw, and ordered him to show cause.
5. Dismissed the plaintiffs' claims with prejudice.
6. Directed the Clerk to send the order to the Oregon State Bar and to Murphy.

The court refused to order money against the plaintiffs themselves, because Rule 11(c)(5) bars monetary sanctions against a represented party for a violation of Rule 11(b)(2). It nonetheless found that Couvrette shared responsibility with her lawyer. It cited unrebutted evidence that she was a serial self-represented litigant who was likely the person who used AI to draft the briefs, and that no one explained what happened. That finding supported the dismissal.

How the $110,204.38 was built

Brigandi's share is $95,998.72. The $500 and $1,000 prices come from the Oregon Court of Appeals, which the federal court said it was adopting. The eight quotations are seven attributed to real cases and one to the Restatement. The court did not add a separate charge for three more fabricated quotations attributed to cases that do not exist.

The Oregon Court of Appeals set that schedule in Ringo v. Colquhoun Design Studio, LLC, 345 Or App 301 (2025) (decided December 3, 2025): $500 per false citation and $1,000 per false quotation or statement of law, which came to $2,000 there. On March 18, 2026, in Doiban v. Oregon Liquor and Cannabis Commission, 347 Or App 742 (2026), the same court applied it to a brief with at least 15 fabricated citations and at least nine fabricated quotations. The formula gave a minimum of $16,500. The court capped the fine at $10,000, citing counsel's detailed explanation, new office procedures, and a response that was filed before Ringo came out.

The federal court also explained why it added fee-shifting on top of the per-item price: "Attorneys weigh risks and rewards for a living." With a known price tag, it reasoned, a lawyer could decide that $500 per fake case or $1,000 per fabricated quotation is outweighed by the reward of outsourcing research to a chatbot.

The local counsel ruling

Murphy did not sign the briefs. His own response said he did a cursory review of the first motion, three days after it was filed, did not review the later briefing, and did not know of the fake authority until defense counsel emailed him on May 5, 2025. The court had warned on December 16, 2024 that further local-rule violations might lead to dismissal and encouraged local counsel to help ensure compliance. In January 2025 Murphy emailed that Brigandi would handle all aspects of the summary-judgment motion, and he later said he would move to withdraw once the motions were fully briefed.

On March 23, 2026 the court held that Murphy's choice to exercise no oversight was a willful violation of Local Rule 83-3. It gave one limit on its own ruling: meaningful participation "does not ordinarily require local counsel to confirm the veracity" of pro hac vice counsel's citations. What it does require, the court said, is more than a bar number and a signature on the pro hac vice application. The court ordered Murphy to attach the order to any future pro hac vice motion he sponsors in the district. It did not ban him from serving as local counsel.

Murphy told the press that he "did not write, review, research, sign or submit the briefs with AI citations" (ABA Journal). That is his account; the March 23 order treated his lack of participation as the violation.

How it compares with other sanction orders

The ByoPlanet row comes from the court's order and its later fee order. In Colorado, the disciplinary judge approved a suspension of one year and one day for Zachariah Crabill, with 90 days to be served and the rest stayed on completion of a two-year probation. The approved summary says he blamed a legal intern when the judge raised concerns and filed an affidavit six days later admitting he used ChatGPT (Presiding Disciplinary Judge, No. 23PDJ067, November 22, 2023).

What courts say about the response after a cite is flagged

Several orders say, in their own words, that what the lawyer did after the problem surfaced affected the sanction. The dollar figures also depend on how many fakes there were and on whether fees were shifted, so read these as examples. They do not add up to a formula.

• Wadsworth: the court listed the drafter's "honesty and candor" as a mitigating fact and appreciated the prompt withdrawal of the motions, the payment of opposing fees and the new policies, though it noted counsel could have acted sooner. It still revoked his pro hac vice status and fined him $3,000.
• Fletcher v. Experian (5th Cir. Feb. 18, 2026): "Had Hersh accepted responsibility and been more forthcoming, it is likely that the court would have imposed lesser sanctions." The sanction was $2,500.
• Lnu v. Blanche (9th Cir. June 3, 2026): the court said the gravity of its discipline, which included a six-month suspension from practice before the court, "is owed to this repeated failure of candor." Its guidance: "There is no upside to denying the use of generative AI or to passing off an AI hallucination as an innocent typographical error."
• Couvrette: the court found no appropriate corrective action and a "coverup" attempt, then used its strongest tools.

In Mata, Judge Castel found bad faith based on "conscious avoidance and false and misleading statements to the Court." The record includes screenshots, attached to Steven Schwartz's May 25, 2023 affidavit, of ChatGPT telling him that Varghese was a real case. The court found his explanations of when and why he asked shifted and contradicted each other, and said that supported its finding of bad faith.

What to do when a cite is flagged

These steps rest on the orders above. They are practice suggestions, not a legal rule.

1. Check every citation and quotation in a primary source, not with the tool that produced them. The Mata record shows ChatGPT confirming that a fake case existed. In Park v. Kim, the Second Circuit said Rule 11 requires, at the very least, that lawyers "read, and thereby confirm the existence and validity of, the legal authorities on which they rely."
2. Tell the court and opposing counsel quickly, and say where the error came from. That is the Lnu instruction, and Wadsworth shows candor counting as mitigation.
3. Fix the proposition, not just the citation. Couvrette treated deleting bad cites while leaving the statements of law in place as inadequate. Lnu said that swapping a fake case for a real one, without saying the first was a hallucination, is not sufficient disclosure. In Doiban, the court allowed a redacted replacement brief only if counsel corrected the description of the five remaining cases, and it required a certification.
4. Do not blame staff. Zachariah Crabill falsely blamed a legal intern when a judge raised concerns about his citations, and his discipline was a suspension. In Farris (6th Cir. Apr. 3, 2026) the court said reliance on "staff" rather than an attorney to supervise AI work product fell short of counsel's obligations.

Three items for a firm policy

Verification before filing. ABA Formal Opinion 512, issued July 29, 2024, applies the existing Model Rules to generative AI. It says lawyers should either acquire a reasonable understanding of a tool's benefits and risks or draw on the expertise of others (Rule 1.1), and that output used in court must be "carefully reviewed to ensure that the assertions made to the court are not false" (Rules 3.1, 3.3 and 8.4(c)). Managerial lawyers must set clear policies and supervisors must make reasonable efforts to ensure compliance (Rules 5.1 and 5.3). The opinion says how much verification is enough depends on the tool and the task. The ABA site blocks automated checks, so open the PDF in a browser. Oregon's own bar guidance, quoted in Ringo, says lawyers must review for accuracy any AI output that discusses case facts or supplies a citation, quotation or conclusion (Oregon State Bar Formal Opinion 2025-205, February 2025).

Client communication. Opinion 512 says a client's informed consent is required before a lawyer inputs information relating to the representation into a self-learning tool whose output could lead to its disclosure, and that general boilerplate in an engagement letter is not sufficient. Whether to record which tool touched which draft is a firm choice; the opinion does not require it.

A named owner for what happens after a flag. The orders above keep returning to how fast and how fully a lawyer responds once a problem is flagged, so decide in advance who runs that response.

For the rules local counsel must follow, read the local rules of the court where you appear. The Couvrette local-counsel ruling applies Local Rule 83-3 of the District of Oregon, and other courts word their rules differently.

FAQ

What was the Oregon AI sanction in Couvrette v. Wisnovsky? A federal magistrate judge in the District of Oregon ordered two lawyers to pay a combined $110,204.38 for summary-judgment briefs that cited 15 nonexistent cases and contained fabricated quotations. The total is a $15,500 fine plus $94,704.38 of the defendants' fees. The court also dismissed the plaintiffs' claims with prejudice.

How much was each lawyer ordered to pay? Stephen Brigandi, the pro hac vice counsel, owes $95,998.72: $15,500 to the Clerk plus $80,498.72 in fees. Timothy Murphy, the Oregon local counsel, owes $14,205.66 in fees for failing to meaningfully participate as local counsel under Local Rule 83-3.

How did the court calculate the $15,500 fine? It used the Oregon Court of Appeals' schedule from Ringo: $500 for each of the 15 nonexistent cases ($7,500) and $1,000 for each of 8 fabricated quotations ($8,000). The court said it adopted that valuation, and it added the defendants' fees because a per-item price alone might not deter.

What rules did the lawyers violate? The court sanctioned Brigandi under Federal Rule of Civil Procedure 11(b)(2) and Local Rule 83-7, which requires lawyers to follow the Oregon Rules of Professional Conduct, including Rule 3.3(a) on false statements to a tribunal. It sanctioned Murphy under Local Rule 83-3 for failing to meaningfully participate as local counsel.

Why was the total so much larger than earlier AI cases? Three things combined: 15 fake cases and 8 fabricated quotations across three briefs, a fee award of $94,704.38 shifted onto the lawyers, and a court that found the amended briefs did not fix the problem. Mata v. Avianca ($5,000) was a flat penalty with no fee-shifting. ByoPlanet also shifted fees and reached $85,567.75.

Has the case ended? No. After the sanctions, the court granted the defendants summary judgment on a breach-of-contract counterclaim on March 30, 2026 and denied reconsideration on June 22, 2026. The plaintiffs appealed to the Ninth Circuit on July 22, 2026 (No. 26-4712), from the summary judgment opinion and the order denying reconsideration according to the docket. On June 12, 2026 the court ordered Brigandi to show cause why he should not be held in contempt for not paying the $15,500. In a July 23, 2026 filing he asked the court to dismiss the sanctions order, said he never used AI, and pointed to his health. The docket shows further filings through October 8, 2026 that we did not read, and we did not read the appeal briefs.

Can a lawyer be sanctioned for using ChatGPT in a brief? The orders punish filing unverified authority. The Ninth Circuit put it this way in June 2026: the rules are violated "at the point of signing and filing," so it generally does not matter whether a fabrication came from an AI tool or from a person.

What should a lawyer do if opposing counsel flags an AI citation? Verify every citation and quotation in a primary source, tell the court and opposing counsel promptly, say where the error came from, and fix the legal propositions that relied on it. Do not ask the chatbot to confirm the cases and do not blame staff.

For the US statutes, regulations and court rules these laws sit in, see the US primary law API.

Arshita Anand

Co-Founder & CEO · Attorney

Arshita leads product and strategy at Vaquill AI, the US primary law data API. As an attorney, she decides which law we carry, how each section is sourced and cited, and what legal AI and compliance teams need from the data.