[EDRM Editor’s Note: The opinions and positions are those of Michael Berman.]
There is nothing new about “hallucinations” from GenAI. A recent decision explains and applies what now appears to be settled law under Fed.R.Civ.P. 11. Brownfield v. Cherokee Co. School Dist., 2026 WL 777230 (E.D. Ok. Mar. 19, 2026). A recent appellate decision applying Fed.R.App.P. 38 and common-law inherent authority is Whiting v. City of Athens, __ Fed.4th __, 2026 WL 710568 (6th Cir. Mar. 13, 2026).
The Whiting court wrote:
The party presentation principle supports this conclusion. “Courts are essentially passive instruments of government” that “wait for cases to come to them” and “rely on the parties to frame the issues for decision.” … And “judges are not like pigs, hunting for truffles buried in the record.” … That all means that we are required to take the cases as we find them. [citations omitted]
See also “Judges are not like pigs, hunting for truffles buried in the record.” (Mar. 31, 2025).
The Whiting court added:
Although citing fake cases violates Federal Rule of Appellate Procedure 38, Rule 38 alone is not “up to the task” of sanctioning this conduct, Chambers, 501 U.S. at 50, 111 S.Ct. 2123, because Rule 38 allows only for the imposition of costs and attorneys’ fees, Sanctions § 33. But we think other sanctions are also appropriate, so we employ our inherent authority.
Our conclusion aligns with that of other courts that have used their “inherent authority to sanction the misuse of AI when it affects the Court’s docket, case disposition, and ruling.”
This blog focuses on Brownfield in which the court applied Fed.R.Civ.P. 11. “Before the Court is the sua sponte order of the Court for Plaintiff to show cause as to why sanctions should not be assessed for the filing of a document containing false or misleading information and citations…. For the reasons stated below, the Court finds that an order of sanctions against Plaintiff under Fed. R. Civ. P. 11 is appropriate.” Plaintiff was pro se.
The facts were straightforward: “Defendants identify eight cases cited in Plaintiff’s Sanctions Motion that either do not exist, did not stand for the legal proposition for which it was cited, do not contain the quoted language, held the opposite of the holding represented by Plaintiff, or has facts and/or an outcome contrary to those represented by Plaintiff.” Plaintiff did not dispute this and admitted to use of GenAI.
The court wrote: “Determining whether sanctions are warranted involves two steps…. The court first must find that a pleading violates Rule 11.… At this step, the person signing the pleading’s conduct is evaluated “under a standard of objective reasonableness.… Second, if the conduct violates Rule 11, a court then imposes an appropriate sanction. Pro se parties are subject to Rule 11 and potential sanctions thereunder…. Finally, because Plaintiff is proceeding pro se, the Court liberally construes his pleadings.” [cleaned up].
The court explained the issue:
“Federal courts increasingly confront filings prepared with the assistance of generative artificial intelligence. While such tools can enhance efficiency, they also create a new professional hazard, synthetic authority presented as precedent.” Mattox v. Product Innovations Research, LLC, 807 F.Supp.3d 1341, 1347 (E.D. Okla. 2025). As the use of AI in legal representation has increased, so have incidences of fictitious or inaccurate citations resulting from such AI tools. Presentation of inaccurate authorities in pleadings resulting from AI subject to the signing party to sanctions. See, e.g., Hill v. Oklahoma, 2025 WL 1840659, at *4 (W.D. Okla. July 3, 2025) (“[C]ourts have increasingly found the imposition of sanctions proper in response to an attorney’s conduct in citing ‘fake cases’ that are ‘hallucinations’ generated by an artificial intelligence platform.”) (citing Wadsworth v. Walmart, 348 F.R.D. 489, 495-496 (D. Wyo. 2025) (citing fake cases in a signed motion violated Rule 11(b)(2) and subjected attorneys to imposition of monetary sanctions); Bunce v. Visual Technology Innovations, Inc., 2025 WL 662398 at *1 (E.D. Pa. Feb. 27, 2025) (imposing sanctions where the attorney filed two separate motions that cited two cases that could not be located “on any legal research tool ….”); Ramirez v. Humala, 2025 WL 1384161 at * 1 (E.D.N.Y. May 13, 2025) (collecting cases finding “the submission of nonexistent case citations in filings to the court to constitute sanctionable conduct” under Fed. R. Civ. P. 11(b)(2)). Plaintiff admits he did not check the case citations generated by AI and, thus, he did not conduct a reasonable inquiry into the law submitted in his Sanctions Motion. Therefore, the Court finds the Plaintiff’s submission of a pleading containing fictitious cases and inaccurate cases references warrants sanctions.
Plaintiff admits he did not check the case citations generated by AI and, thus, he did not conduct a reasonable inquiry into the law submitted in his Sanctions Motion. Therefore, the Court finds the Plaintiff’s submission of a pleading containing fictitious cases and inaccurate cases references warrants sanctions.
Brownfield v. Cherokee Co. School Dist., 2026 WL 777230 (E.D. Ok. Mar. 19, 2026).
It reasoned:
Acknowledging his use of AI assisted legal research in the Sanctions Motion, Plaintiff took the affirmative remedial step of completing a two-month AI program costing $1,800 that covered AI hallucinations and ethical issues prior to the show cause hearing…. Plaintiff further pledged not to use any generative AI tool in any future matter before this Court…. While the Court commends Plaintiff for acknowledging his use of AI and taking responsibility for the inaccurate citations resulting therefrom, the Court must still consider the White factors in assessing an appropriate sanction. The saliant factors under the circumstances of this case are punishing present litigation abuse and compensating the victims of the litigation abuse…. Of note, Defendants’ counsel spent time searching for and analyzing the authorities cited in the Sanctions Motion. Although not broken down specifically in Mr. Rainey’s affidavit, it is reasonable to conclude that Defendant’s attorneys incurred time attempting to locate the authorities cited and reviewing those cases for the propositions for which Plaintiff cited them.
A $500.00 sanction was imposed. “This represents approximately two hours of Ms. Albers billable hourly rate in researching and identifying such citations. Furthermore, this amount serves as both punishment for the presentation of such citations and compensates Defendant for legal expenses resulting therefrom. In conjunction with Plaintiff’s representation of himself pro se in this matter, his candor and proactive remedial measures, and his pledge not to make further use of AI in this case, the Court believes this sanction is appropriate as the least severe sanction to adequately deter and punish Plaintiff’s conduct.”
Assisted by GAI and LLM Technologies per EDRM’s GAI and LLM Policy.
