Lawyers cite fake AI cases
- Scientific American reported on May 23 that lawyers are still filing AI-invented case citations in court despite repeated judicial warnings and sanctions. - An Alabama Supreme Court opinion on April 24 called one lawyer’s briefs “grossly deficient” and sanctioned him after fake citations kept appearing. - Courts’ next guidance is likely to come in sanction orders and local filing rules, including future Alabama and federal cases.
Scientific American reported on Saturday that lawyers are still submitting court filings with case citations invented by generative AI, nearly three years after a federal judge in New York turned one such episode into a national warning. The article pointed to a growing list of sanctions, dismissed filings and show-cause orders in state and federal courts. Judges have responded by fining lawyers, striking briefs and, in some cases, imposing training or other conditions on future filings. The pattern has shifted the issue from a one-off embarrassment to a recurring problem in routine litigation. ### Which case made this problem impossible for judges to ignore? June 22, 2023, is the date most lawyers now associate with the issue. In *Mata v. Avianca*, U.S. District Judge P. Kevin Castel of the Southern District of New York sanctioned lawyers after a filing included non-existent judicial opinions, fake quotes and fake citations created by an AI tool. The court imposed a $5,000 sanction. The *Mata* opinion mattered because it showed the mechanics of the failure in detail. Judge Castel wrote that after opposing counsel and the court raised questions, one attorney still submitted material generated by ChatGPT that described cases that did not exist. The order turned “AI hallucinations” from a tech term into a courtroom problem with a citation every litigator could read. (national.clla.org) ### If lawyers know about Mata, why are fake citations still showing up? May 23, 2026, Scientific American’s article said the repeat offenses suggest the problem is no longer simple unfamiliarity with generative AI. The magazine wrote that attorneys keep delegating research and drafting tasks to systems that can invent authorities, then fail to verify the output before filing it. (national.clla.org) May 28, 2025, offers one recent example. Bloomberg Law reported that Judge James Patrick Hanlon of the Southern District of Indiana imposed a $6,000 sanction on an attorney who filed briefs with nonexistent cases generated by AI. Magistrate Judge Mark J. Dinsmore wrote in recommending a higher penalty that earlier sanctions had “evidently failed to act as a deterrent.” He added that confirming a case exists and remains good law is “a basic, routine matter” for a practicing attorney. (scientificamerican.com) ### What are courts doing when they catch it? April 24, 2026, the Alabama Supreme Court dismissed an appeal in *Ibach v. Stewart* after finding that attorney W. Perry Hall had filed briefs containing “numerous invalid, inaccurate, and/or irrelevant citations” that appeared to be AI “hallucinations.” The court said the briefs were “grossly deficient” under Alabama’s appellate rules and said Hall’s conduct was “so egregious as to warrant sanctions.” (news.bloomberglaw.com) February 24, 2025, a federal judge in Wyoming sanctioned lawyers from Morgan & Morgan and the Goody Law Group after a motion cited eight nonexistent cases, according to the ABA Journal’s report on the order. U.S. District Judge Kelly H. Rankin had first directed the lawyers to produce the cited authorities or explain why they should not be sanctioned. The firms said the cases had been “hallucinated” by an internal AI platform and were not legitimate. (publicportal-api.alappeals.gov) ### Why does this keep becoming a sanctions issue instead of just a bad-research issue? Rule 11 in federal court and comparable state rules already require lawyers to make reasonable factual and legal inquiries before filing papers. In *Mata*, Judge Castel said sanctions were warranted because the filing misrepresented authorities to the court. In Alabama, the state supreme court tied the problem directly to briefing rules and its inherent authority to manage proceedings before it. (abajournal.com) Scientific American framed the broader concern as institutional reliance. Courts generally depend on lawyers’ briefs to accurately identify authorities, and that system breaks down when fabricated citations are inserted into filings that judges and clerks must then verify one by one. ### What should readers watch next? Future court orders will likely provide the clearest measure of whether sanctions are changing behavior. (national.clla.org) Alabama’s April 24 opinion, the Indiana sanction order entered on May 28, 2025, and the Wyoming sanctions issued on February 24, 2025, show judges are building a record of escalating responses. New local rules, certification requirements and show-cause orders in pending cases will be the next places to watch named judges and courts address AI-generated legal citations. (scientificamerican.com) (publicportal-api.alappeals.gov)