The court found an AI program generated the fabricated citations and rejected the lawyer's LexisNexis explanation.
This was a dispute in the Southern District of New York where Sai Malena Jimenez-Fogarty sued Thomas Fogarty and others, including two judges. Her attorney, Tricia S. Lindsay, filed two memoranda of law opposing motions to dismiss. Both briefs cited numerous cases that do not exist.
The court walked through the fabrications one by one. Lindsay cited "Cornhill LLC v. Sowers, 183 A.D.3d 649," but the reporter pointed to an unrelated case, and no case by that name could be found. She cited a "Bank of Am., N.A. v. Gruff" decision the court could not locate, plus several others whose citations led to entirely different cases on unrelated topics.
Opposing counsel for two of the defendants first flagged the bad citations in a reply brief. The court then issued orders to show cause. Lindsay never explained where the citations came from, at one point suggesting LexisNexis AI features "may very well have" generated them, a claim the court rejected for lack of any evidence.
Offered for: a forfeiture proposition
The court found that Lindsay violated Rule 11, concluding she used an AI program that generated the fabricated citations and failed to check whether they were real. Finding clear and convincing evidence of bad faith, the magistrate judge sanctioned her $2,500. The court also noted she filed more false citations on two later occasions and was referred to the Second Circuit's Grievance Panel in a related appeal. It ordered her to pay the fine, give her client a copy of the order, and notify the judge in every pending case where she appears.
Kingsfield returned REJECT. The identifier and party name match no opinion on the cited topic. The citation pointed to an unrelated docket.
This matter is among the US case-law cases we ran Kingsfield against in the Charlotin benchmark. Kingsfield's existence check confirms every citation against the case-law corpus and an independent source (CourtListener) before any claim resting on it can stand. Across the run Kingsfield ruled 696 of 700 testable claims correctly, including every one of the 504 fabricated citations and 92 fabricated quotes.
Honest scope. This measures Kingsfield's ruling layer given an extracted citation, not an end-to-end run over a full brief. The case record and the AI-tool attribution come from Damien Charlotin's public database and the court order; the verdict is Kingsfield's. See the full 700-claim benchmark.