Mata v. Avianca / Fabricated Citations
A 2023 U.S. District Court case (S.D.N.Y., Judge P. Kevin Castel) in which plaintiff's attorneys submitted a brief built from ChatGPT-fabricated case citations and were sanctioned $5,000 for bad-faith conduct. The case became the canonical example of LLM citation hallucination causing real legal consequences and has been followed by hundreds of similar sanctions across U.S. courts.
Mata v. Avianca, Inc. (1:22-cv-01461, S.D.N.Y.) began as a routine personal-injury suit: plaintiff Roberto Mata alleged a metal serving cart struck his knee on an Avianca flight in 2019. The case became notable for what happened in the briefing. After Avianca moved to dismiss, the plaintiff's lawyers submitted an opposition brief citing several decisions — *Varghese v. China Southern Airlines*, *Shaboon v. Egyptair*, *Petersen v. Iran Air*, and others — that did not exist. Avianca's counsel could not find the cases; the court could not find them either. Ordered to produce copies, attorney Steven Schwartz of Levidow, Levidow & Oberman attached purported excerpts that were themselves nonsense. Schwartz had used ChatGPT to draft the brief and had asked the chatbot whether the cited cases were real; ChatGPT had assured him they were available on Westlaw and LexisNexis. He later told the court he believed ChatGPT was a 'super search engine' rather than a generative model that could fabricate references. Co-counsel Peter LoDuca signed the filings without verifying them. On June 22, 2023, Judge P. Kevin Castel sanctioned Schwartz, LoDuca, and the firm $5,000 jointly for bad-faith conduct — emphasizing not the use of AI itself but the attorneys' failure to investigate after the fake-case problem surfaced. He also required them to send the order to each judge falsely identified as the author of one of the hallucinated opinions. Mata's underlying claim was dismissed on statute-of-limitations grounds. The case has become the canonical reference for LLM hallucination in professional practice. A public tracker maintained by legal researchers has logged hundreds of subsequent incidents in U.S. federal, state, and tribal courts through 2025, including a California sanction described as the largest state-court penalty for attorney AI misuse and a 2025 ruling in Johnson v. Dunn that disqualified attorneys from a case after similar fabrications. Bar associations across the U.S. and several other jurisdictions have since issued AI guidance citing Mata as the cautionary example.