Use Of AI-hallucinated Cases Results In $100K In Penalties And Dismissal Of Action (Couvrette v. Wisnovsky)

Green Bldg. Initiative, Inc. v. Peacock, WL 3002143

Notwithstanding the warnings against the reckless use of generative artificial intelligence to conduct legal research or draft factums or other court documents, lawyers, clients, and self-represented litigants continue to do so at their own peril. Lawyers who either assume the responsibility of filing court documents under their name or who have obligations under court rules in some jurisdictions to review those documents must take great care to ensure that non-existent cases or fabricated quotations are not used. The failure to do so can result in significant financial penalties. Similarly, clients and self-represented litigants who want to save a “buck or two” on legal fees, must also be wary that their decisions to file court documents that have been drafted using generative AI can result in the dismissal of their claim or defence in circumstances where those documents contain fakes cases or hallucinated propositions of law.

In Couvrette v. Wisnovsky, Case No. 1:21-cv-00157-CL (United States District of Court, District of Oregon, United States Magistrate Judge Clarke), the court ordered over $100,000 in penalties against two lawyers for the citation of several non-existent cases and fabricated quotations in three court filed briefs. The plaintiffs’ action was also dismissed on the basis that the individual plaintiff, J, had allegedly prepared the briefs.

This case involved a family dispute over a winery. The plaintiffs had commenced the action against the defendants, who were J’s brothers, in connection with tenancy and ownership issues over the winery and elder abuse issues. The defendants made a counterclaim against the plaintiffs.

Although the action was commenced in Oregon, the plaintiffs wanted to use a California lawyer to litigate the action. However under the local court rules of Oregon, the plaintiffs were still required to engage a local lawyer, who was obligated to “meaningfully participate” in the matter. While the plaintiffs did so, in 2024, without the knowledge of the local lawyer, the plaintiffs commenced two small claims court actions against the same parties over the same property, and filed a motion in violation of the local rules.

In response to the motion, the court gave notice that “[n]oncompliance with Local Rules is not something this Court takes lightly, and further noncompliance may result in denial or dismissal of this case.” This notice would later be used by the court as an alternative ground to dismiss the plaintiffs’ action.

On January 31, 2025, the plaintiffs brought a summary judgment motion against the defendants. In this motion, the plaintiffs cited two non-existent cases.

At the same time, the defendants brought a cross-motion for partial summary judgment.

On April 4, 2025, the plaintiffs filed responding materials to the cross-motion and the defendants filed materials in response to the plaintiffs’ motion. The defendants’ materials identified the two non-existent cases in the plaintiffs’ initial materials and contended that local rules had been breached.

At this time, the defendants did not know that the plaintiffs’ responding materials cited seven more non-existent cases.

Three weeks later, the plaintiffs filed a “Notice of Errata” to correct the errors concerning the two non-existent cases.

On May 2, 2025, the plaintiffs then filed a reply brief in support of their summary judgment motion. This brief cited six more non-existent cases and contained false quotations from seven real cases and a fabricated quotation from the Restatement (Second) on Contracts.

The plaintiffs’ use of non-existent cases and their breaches of the local rules caused the defendants to bring a motion for sanctions under applicable US Federal Rules of Civil Procedure. While this motion caused local counsel to bring his own motion to withdraw from the case and some additional procedural wrangling, the court eventually heard a motion for sanctions to determine the appropriate penalties and relief for the plaintiffs’ misuse of generative AI and continual violations of the local court rules.

In two separate rulings dated December 12, 2025 and March 23, 2026, the court assessed the penalties that should be imposed against the California lawyer, the local lawyer and the plaintiffs for the misuse of generative AI.

Overall, the court found that the plaintiffs and their lawyers had failed to acknowledge, explain or appropriately amend the portions of their initial brief which contained the two non-existent cases and that no meaningful explanation had been provided or corrective action taken throughout the litigation for the repeated use of fake cases and fabricated quotations.

With respect to the California lawyer, the court specifically found that he had provided no evidence of any steps taken to verify the veracity of legal arguments made in the briefs. The court noted that he had also failed to accept responsibility for the use of the non-existent cases and fabricated quotations when responding to the defendants’ lawyers, who advised that they would bring a sanctions motion for the misuse of generative AI. Instead, in response, the California lawyer wrote, in part: “Plaintiffs will not agree to the sanctions requested. If such Motion is filed, Plaintiffs will respond with a Notice of Errata and Corrections.”

While the plaintiffs filed a second Notice of Errata in response to the sanctions motion, the court found that this Notice simply contained superficial amendments, such as the removal of fabricated quotations, without any changes to the surrounding text, or the removal of quotation marks from around purported statements of law without the removal of the false statements of law in their entirety.

Accordingly, in the December ruling, the court ordered the California lawyer to pay the court $15,500 for submitting 15 non-existent cases and 8 fabricated quotations falsely attributed to legitimate cases in three briefs. This penalty was based on $500 per non-existent case and $1,000 per fabricated quotation.

With respect to the plaintiffs, the court found that they shared responsibility with the California lawyer for the submission of non-existent citations and fabricated quotations. There was no evidence from J that she was not responsible for drafting the briefs or that she had any reason to believe that the California lawyer was verifying her drafts prior to submission to the court.

Furthermore, the plaintiffs had been warned about breaching the local rules and the court’s authority to dismiss their claims. In the circumstances, the court found the severity of the breaches and misuse of AI warranted the dismissal of the plaintiffs’ claims.

With respect to the local lawyer, he sought to avoid being sanctioned on the grounds that he had not participated in the use of non-existent cases or fabricated quotations and that he was unaware that the local court rules required him to review the California lawyer’s court briefs before submission to the court. He further submitted that he did not act willfully in failing to “meaningfully participate” in the matter or in bad faith.

However, the local rules were clear, and, like the plaintiffs, the local lawyer was warned that the court did not take a breach of the local rules lightly. The court noted that the Oregon Chapter of the Federal Bar Association had advised that lawyers should err on the side of participating in a meaningful manner in cases of doubt. By choosing not to meaningfully participate, the local lawyer left the court to muse that it would never be known what could have been prevented if he had complied with the local rules. In the circumstances, the court found that the local lawyer willfully violated the rule regarding meaningful participation.

In the December ruling, the court concluded that the defendants were entitled to reasonable fees and costs directly resulting from the misuse of generative AI. Since the Federal Court Rules did not permit monetary sanctions against a represented party, the lawyers bore the brunt of this award. Accordingly in the March ruling, the court awarded fees and costs in the amount of $94,704.38 against the plaintiffs’ lawyers.

The local lawyer was ordered to pay 15% of this award for his failure to meaningfully participate in the case in willful disobedience of local court rules. The California lawyer was ordered to pay the remaining 85%.

The key takeaway from this case is that the penalties for the misuse of generative AI can be significant and that to avoid hefty monetary sanctions and the potential dismissal of a claim or a defence, lawyers, clients and self-represented litigants must pull their heads out of the sand and stop blindly relying on the use of generative AI to prepare legal arguments and briefs. Every lawyer should be on alert to the misuse of generative AI and do everything possible to verify all citations, quotations and propositions of law cited in a document before it is submitted to court.

Although Courvette is a US case, in Canada, rules of civil procedure and/or practice directions permit courts to award costs personally against lawyers for misconduct during a proceeding, and to dismiss actions or defences for the misuse of AI.

Furthermore, although this US case is not directly on point, it may by analogy be informative about the role opposing lawyers in all jurisdictions must play to ensure that non-existent cases or fabricated quotations and propositions of law are never argued before a judge or tribunal. As officers of the court, all counsel are obligated to ensure that a court is never misled. In this regard, Judge Miller in Nuvola, LLC v. Wright, Court File No.: 27-CV-HC-15-3802 (Minnesota District Court, County of Hennepin, Fourth Judicial District, November 21, 2025), chastised opposing counsel for his failure to bring non-existent citations in the opposing party’s brief to the court’s attention as follows:

The Court also finds troubling Mr. Braun’s failure to identify or bring the non-existent case citations to the Court’s attention before the hearing on the motion to compel arbitration. The Court should not be left as the last line of defense against citations to fictional cases in briefs filed with the court. While Mr. Braun did not create or rely on the fake citations, he also did not detect them. Instead, he admitted he did not review the cases citated by his opponent. If he had checked out the citations in the brief to which he was responding, he no doubt would have brought the issue to the Court’s attention by the time of the motion hearing, and that would have allowed the Court to take the non-existence of the cited cases into consideration as it heard the argument on the merits of Defendant’s motion to compel arbitration, instead of leaving the Court to discover that issue on its own, after the hearing was concluded. The Court does not find Mr. Braun’s conduct to be sanctionable, as he did not cite any non-existent cases to the Court. Nonetheless, the Court reminds counsel that it is the obligation of counsel on both sides to respond to each other’s arguments, including completing a basic cite-check of the cases cited by the other side.

[This blog was written without the use of generative artificial intelligence. For Canadian lawyers who are using generative AI to draft legal arguments or relying on clients or third parties to do so, as a research lawyer, I offer my services to check your documents to identify non-existent cases and fabricated quotations prior to submitting them to court.] A PDF version is available to download here.

The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.