SIDEBAR

• A federal judge entered default judgment against Affable Avenue LLC after its attorney repeatedly filed papers containing nonexistent cases and falsely attributed quotations.
• Another defendant’s lawyer uncovered at least 13 nonexistent cases and eight real cases containing quotations they did not actually say.
• The attorney has now been ordered to pay $38,877.59 to the law firm whose lawyers spent time investigating, reporting and litigating the citation misconduct.

By Samuel López | USA Herald

NEW YORK — A federal lawyer’s failure to verify AI-generated legal research did far more than produce an embarrassing correction. It ultimately cost his client its defense in the case.

U.S. District Judge Katherine Polk Failla sanctioned attorney Steven A. Feldman after finding that he repeatedly submitted false legal citations while representing Affable Avenue LLC in Flycatcher Corp. Ltd. et al. v. Affable Avenue LLC et al., Case No. 1:24-cv-09429-KPF, in the Southern District of New York. The Judge’s order included monetary sanctions and case-terminating sanctions.

The underlying lawsuit concerns alleged trademark infringement and unfair competition involving children’s toys, along with claims involving deceptive trade practices, the alleged sale of stolen goods and related unfair competition.

Case-terminating sanctions means exactly what it sounds like. Rather than merely warning or fining a lawyer, the court imposes a sanction that effectively ends that party’s ability to contest the case. Here, Failla struck Affable’s filings and entered default judgment against Feldman’s client. That means the lawyer’s conduct had a direct and severe consequence for a client that was defending the lawsuit.

The trouble began with Affable’s June 2025 motion to dismiss. Joel MacMull, then representing co-defendant Top Experience Company LLC, examined Feldman’s authorities and reported at least 13 cases that did not exist and eight real cases that did not contain quotations attributed to them. MacMull first contacted Feldman privately on June 22, 2025, identifying the errors and asking him to withdraw and correct the filing. When that did not happen, MacMull notified the court.

One later example is striking. Feldman cited “Himmelstein v. Comcast of the D.C., LLC, 908 F.3d 49 (D.C. Cir. 2018), aff’g 44 F. Supp. 3d 1 (D.D.C. 2012).” The court determined those authorities, as cited, did not exist. The reporter pages actually pointed to unrelated decisions. The case Feldman apparently intended to cite was Himmelstein, McConnell, Gribben, Donoghue & Joseph, LLP v. Matthew Bender & Co., Inc., 37 N.Y.3d 169 (2021).

There was another problem. In responding to the court’s scrutiny, Feldman appeared to attribute language to Mata v. Avianca Inc., the well-known earlier AI hallucination case. Judge Failla found that the quoted language was not in Mata at all. A Google search instead traced it to a 2023 law-firm article discussing Mata.

Feldman acknowledged using research tools including vLex and Paxton AI and admitted he had not personally completed a full cite-check before filing. The judge concluded AI had generated nonexistent authorities and misattributed quotations that were never properly verified.

The newest financial sanction shows another side of federal judicial power.

Rule 11 does not require every monetary sanction to be paid into the U.S. Treasury or to a client. Depending on the legal basis and circumstances, a court can require payment of reasonable attorney fees and expenses caused by misconduct. Here, Failla had specifically authorized MacMull to seek fees after finding Feldman acted in bad faith and had unreasonably multiplied the proceedings.

The court awarded $38,877.59 to MacMull and his firm, Mandelbaum Barrett PC, reducing the requested fees by 10%. Failla noted that MacMull had initially tried to resolve the fee issue directly, but said Feldman gave him the “runaround.” The court ordered Feldman to pay within 30 days unless the parties agree to another schedule.

The lesson for lawyers using AI is not that courts prohibit the technology. Failla expressly said that was not the problem. The problem was filing authorities without reading and verifying them — and continuing after warnings.

About the Author

Samuel López is a USA Herald reporter and correspondent with more than two decades of experience working in the legal sector, including civil litigation, legal research, complex motion practice and analysis of state and federal court proceedings. His legal background informs his reporting on artificial intelligence in law, litigation ethics, court sanctions and the growing responsibility lawyers face when incorporating generative AI into professional legal work.

Read Beyond the Headlines

If you value legal reporting that goes beyond the press release and examines the court filings, legal consequences and facts behind the story, join the USA Herald newsletter for exclusive investigative reporting and legal analysis.

👉 Follow USA Herald on X: https://x.com/RealUSAHerald