New Mexico lawyer fined over ChatGPT-fabricated testimony

New Mexico's Supreme Court has fined attorney Stephen Aarons $5,000 after he submitted an appeal brief containing testimony from witnesses invented by ChatGPT. The sanction was reported by The Verge on 11 September 2026, following the court's order that week. The case concerns the appeal of Oscar Renee Sandoval, who was convicted of murder and sentenced to life in prison.

The court's written order describes four wholly fabricated witnesses, including two identified as police officers. It also lists false statements attributed to other witnesses about threats and the shooter's clothing and appearance. The brief misrepresented existing legal authorities as well. Aarons admitted that he had used ChatGPT to prepare the document and had not checked its factual claims or legal authorities before signing and filing it.

The consequences extend beyond the fine. The court found Aarons in direct contempt, referred the matter to its disciplinary board and barred him from appearing before the state's Supreme Court pending that process. It appointed the public defender's office to take over the appeal, struck the previously filed briefs and ordered a fresh briefing schedule once new counsel appeared. The intention is to hear the appeal during the court's 2026–2027 term.

According to reporting cited by The Verge, Aarons told an August hearing that he had supplied ChatGPT with a computer-generated trial transcript and other case materials, expecting an accurate summary. He subsequently described the episode to Reuters as an honest mistake and expressed remorse. The court's order, however, records concern about his response and his failure to inform his client about the inaccurate filing.

For professional AI users, the practical lesson is that summarising supplied documents still requires checking the result against those documents. A plausible account can introduce people, quotations or events that the underlying record does not contain. In this case, the errors reached a proceeding affecting someone's liberty.

For organisations adopting AI, the case illustrates why review needs to cover factual assertions as well as citations. Keeping original records accessible and assigning responsibility for verification are concrete safeguards. The court's decision addresses this lawyer's conduct; it does not establish that every use of AI in legal work is prohibited.