The Rising Cost of AI-Hallucinated Citations

Since ChatGPT debuted in late 2022, attorneys and pro se litigants alike have taken to outsourcing legal research and writing to generative AI. But whereas outsourcing in many spheres escapes regulation, not so in the law. Procedural rules and ethical duties across jurisdictions require legal filings created with AI’s assistance to be vetted and purged of errata. Penalties for plying courts with AI-generated fake facts and counterfeit caselaw can be severe.

The latest hefty sanction was meted out by the Illinois Appellate Court, First District, in Scott v. Illinois Human Rights Comm’n, 2026 IL App (1st) 251462. Appellant’s counsel was fined $15,000 and reported to the ARDC for his AI-hallucinations. Id., ¶¶ 56, 82. Counsel’s opening and reply briefs contained:

Id., ¶¶ 20–28, 56.

When ordered to show cause, counsel, to his credit, acknowledged the suspect citations were in fact false. Id., ¶ 37. He “‘frequently use[s] a premier corporate subscription to ChatGPT’ to ‘synthesize complex matters,’” the case at bar included. Id. Despite personally verifying every citation via Lexis-Nexis, the ones at issue had eluded him. Id. Counsel ascribed his “‘inability to properly verify the cites’” to his inexperience with appellate work. Id. He suggested a fine of $1,000 for the non-existent case and $500 for each of the remainder. Id. Unfortunately, counsel’s response also contained misstatements of law, prompting the court to condemn it as “yet another sloppy filing.” Id., ¶¶ 38, 41. The court went on to lament that this “type of lawyering is harming . . . clients, his own professional reputation, and this court’s workflow.” ¶ 41.

Like Illinois’s first ruling on AI-misuse—In re Baby Boy, 2025 IL App (4th) 241427—on which the Scott court drew heavily, sanctions were administered under Supreme Court Rule 375. Id., ¶¶ 34, 43. That Rule gives discretion to sanction those who (a) willfully fail to abide by the Rules or (b) bring a frivolous appeal. Ill. S. Ct. R. 375. The Scott court ruled counsel willfully violated Rule 341(h)(5) because he doubled down in his reply brief on the sham statutory law in his opening brief. Id., ¶ 43. And since his arguments derived from a hoax (a non-existent statutory requirement), his arguments were frivolous. Id., ¶ 44.

The court considered additional non-monetary sanctions, including ordering counsel to serve a copy of the opinion upon every judge before whom he has a pending matter and ordering him to attend CLE courses. Id., ¶¶ 60–61. But uncertain whether Rule 375 allows for them, the court refrained. Id., ¶ 62. It did report counsel to the ARDC for professional misconduct. Id., ¶¶ 45, 59.

While the arguments were frivolous, the court noted the appeal was not. Id., ¶ 44. The plaintiff was a woman appealing the Illinois Human Rights Commission’s dismissal of her workplace racial discrimination charge. Id., ¶ 1. Despite counsel’s misconduct, the court declined appellees’ request to strike the briefs and dismiss the appeal. Id., ¶ 18. Citations to fictitious legal authorities, the court stressed, do not result in automatic dismissal or striking of a brief. Id., ¶ 33 (citing Baby Boy, 2025 IL App (4th) 241427, ¶ 126). The record on appeal and the pertinent statutory law enabled the court to affirm the Commission’s dismissal of appellant’s charge for lack of jurisdiction. Id., ¶¶ 1, 33.

The court ordered a fine of $1,500 per transgression. Id., ¶ 56. This rate was higher than what courts had previously imposed, and the court offered two reasons for it. Id., ¶ 57. First, a flat, as opposed to a variable, rate was in order because “there is no meaningful difference between the various types of false citations and quotations.” Id., ¶ 57. Second, “AI-hallucinated citations are an increasingly persistent problem in courts nationwide,” indicating that current fines do not deter AI-misuse. Id., ¶ 58. Thus, courts “have no choice but to increase fines . . .  until those fines have a significant deterrent effect.” Id. The court rejected the notions “that AI hallucinations are just ‘technical deficiencies’” or an unavoidable occupational hazard of modern practice. Id., ¶ 48.

This reasoning opens the prospect of even greater fines. But should fines not stifle misconduct or hit a ceiling, what then? So far, no practitioner has been disbarred for filing a brief contaminated by AI-hallucinations. But courts are already issuing temporary suspensions. See Malkeet Lnu v. Blanche, 177 F.4th 1014 (9th Cir. 2026) (suspending counsel from appearing before the court for six months). Time will tell if suspension sanctions mirror the upswing in monetary sanctions.

Author: Vitaliy Oprysko is a recent graduate from DePaul University School of Law. He is a clerk with the Keleher Appellate Law Group and has worked on appeals in the Illinois Appellate Court and the U.S. Courts of Appeals for the Sixth and Seventh Circuits.