> ## Content Index > Fetch the complete content index at: https://www.tao.media/llms.txt > Use this file to discover other available public pages before exploring further. # New Mexico Supreme Court Fines Lawyer After ChatGPT Invented Witnesses in Murder Appeal - URL: https://www.tao.media/new-mexico-supreme-court-fines-lawyer-after-chatgpt-invented-witnesses-in-murder-appeal/ - Published: 2026-09-12T15:37:00.000Z - Updated: 2026-09-12T15:37:00.000Z - Description: The court struck the filings, pulled the lawyer off the case, and sent them to discipline. - Author: Bart Hillerich - Tags: AI, News The [New Mexico Supreme Court](https://www.nmcourts.gov/supreme-court/?ref=tao.media) held Santa Fe defense lawyer Stephen D. Aarons in direct contempt and fined him $5,000 after he filed an AI-generated appeal brief that included witnesses and testimony that did not exist. The case centers on the appeal of Oscar Renee Sandoval, who was convicted of murder and sentenced to life in prison. Aarons, a criminal defense lawyer with more than 40 years at the bar, had been hired by Sandoval's family to handle the appeal. Instead, the state's highest court found that his brief misrepresented the trial record, attributed statements to real witnesses who did not give them, and cited testimony from people the court said were "wholly fabricated." The order, first covered nationally by outlets including [Ars Technica](https://arstechnica.com/tech-policy/2026/09/chatgpt-using-lawyer-punished-for-citing-fake-testimony-from-made-up-witnesses/?ref=tao.media) and [The Guardian](https://www.theguardian.com/technology/2026/sep/11/new-mexico-lawyer-ai-chatgpt-testimony?ref=tao.media), is a sharper version of a now-familiar AI problem in courtrooms. Since the 2023 *Mata v. Avianca* sanctions order brought national attention to fake AI-generated case citations, courts have repeatedly warned lawyers that generative AI output must be checked before it is filed. The New Mexico matter, however, goes beyond bogus citations. The disputed brief inserted invented people and statements into a criminal appeal involving a life sentence. ## What the Court Found The court said Aarons filed an AI-generated brief without verifying either its factual claims or its legal authority. He admitted that he signed and filed the brief without checking the AI output against the record and without telling his client that the filing contained factual and legal misrepresentations. The sanction includes a $5,000 fine payable to the State Bar of New Mexico Client Protection Fund. The court also referred Aarons to the Disciplinary Board, barred him from appearing before the New Mexico Supreme Court until that process is complete, removed him from Sandoval's case, and ordered new counsel to be appointed. According to [Ars Technica](https://arstechnica.com/tech-policy/2026/09/chatgpt-using-lawyer-punished-for-citing-fake-testimony-from-made-up-witnesses/?ref=tao.media), the Supreme Court also struck prior briefs from the record and said the appeal would proceed in the court's 2026-2027 term. Public defender Kim Chavez Cook was assigned to represent Sandoval after Aarons was removed. The court's order said Aarons "demonstrated a lack of remorse and a lack of concern for his client." That client-facing concern became a major issue at an Aug. 21 hearing, where justices questioned why Aarons had not fully explained to Sandoval that the brief contained AI-generated falsehoods. ## ChatGPT Invented Witnesses and Testimony Aarons told the court that he uploaded a computer-generated trial transcript and other case materials into ChatGPT and asked the system to produce what he believed would be a "bulletproof summary." The output was not. The court said the brief included testimony from Officer Michelle Amarillo, Officer Sanchez, Manal Al-Jibury, and Teresa Marquez, even though those witnesses were fabricated. The filing also attributed statements to real people that the court said they did not make. Danny Stanton was described as having received threats from Sandoval, and Linda Stanton was described as testifying about those threats. The court said that testimony was false. The brief also included statements about the shooter's clothing and appearance from Mariah Chavez and Marquez, including a detail that the shooter wore "dark pants and a white shirt." The brief also misrepresented legal authority from at least two New Mexico cases. Aarons later emphasized that he had not cited entirely fake cases. Justice C. Shannon Bacon rejected that distinction during the hearing, saying the duty of candor applies whether a lawyer invents a case or inaccurately describes a real one. For the trial on hand, these errors matter because an appeal brief is supposed to frame the record for legal review. A court does not retry the case from scratch; it reviews arguments based on what happened in the trial court. If a brief imports facts that were never in the record, it can distort the legal questions before the appellate court and impose new costs on the opposing party, the court, and the defendant whose appeal must be rebuilt. ## Why This AI Court Sanction Stands Out The legal profession has spent the last several years absorbing warnings about AI hallucinations. Many sanctions cases have involved nonexistent cases, fake quotations, or inaccurate descriptions of precedent. Those are serious problems, but they usually concern the legal authority section of a brief. The New Mexico order shows a different risk. Generative AI can also fabricate the underlying factual record. In a criminal appeal, that is especially consequential. The disputed filing was not part of a low-stakes procedural fight. It concerned the appeal of a murder conviction and a life sentence. Justice Bacon made that point directly at the Aug. 21 hearing. "Counsel, do you watch the news? Do you listen to the radio? Do you read anything about what's going on in the world?" she asked, according to reporting from [The Verge](https://www.theverge.com/ai-artificial-intelligence/994207/chatgpt-new-mexico-lawyer-fined-murder-appeal?ref=tao.media). "Because the problem with lawyers relying on AI hallucinations is an above-the-fold story every single day." Aarons told Reuters and other outlets that he did not understand how badly the model could hallucinate facts and described the episode as an honest mistake. At the hearing, he also called it "stupidity." The court's response was that tool choice did not change the lawyer's responsibility. ChatGPT did not file the brief; Aarons signed it.