SANTA FE
The New Mexico Supreme Court held attorney Stephen Aarons in direct contempt, imposing a $5,000 fine and referring him to disciplinary review for submitting a brief containing "false testimony from wholly fabricated witnesses" generated by ChatGPT.
Aarons, a criminal defense lawyer with over 40 years of practice in New Mexico, admitted he did not verify factual claims or legal citations before signing and filing the brief. He also failed to inform his client, Oscar Renee Sandoval, of the omissions or the brief's misrepresentations.
The August 2025 filing, submitted as part of an appeal for Sandoval's murder conviction, contained fabricated testimony from Officer Michelle Amarillo, Officer Sanchez, Manal Al-Jibury, Teresa Marquez, Danny Stanton, Linda Stanton, and Mariah Chavez. The false quotes ranged from invented threat allegations to fabricated witness descriptions of the shooter. The brief also mischaracterized legal precedent.
At an August 21 hearing, Aarons testified that he fed a computerized transcript of the trial and case documents into ChatGPT, using the OpenAI o3 model released earlier in 2025. He said he "assumed that it generated a bulletproof summary of proceedings" and relied on the AI's widespread adoption as evidence of reliability.
Justice C. Shannon Bacon confronted Aarons' lack of due diligence directly: "Counsel, do you watch the news? Do you listen to the radio? Do you read anything about what's going on in the world? Because the problem with lawyers relying on AI hallucinations is an above-the-fold story every single day."
The court's order concluded that Aarons "demonstrated a lack of remorse and a lack of concern for his client." Aarons himself told the court, "It's of little comfort to know that my stupidity is what brings us together this afternoon."
The case exposes a gap between AI capability and professional liability. Legal services operate on razor-thin error margins—misquoted precedent or fabricated testimony can nullify appeals, sanction attorneys, and leave clients in prison. Where a consultant's AI recommendation might cost a business efficiency, a lawyer's hallucinated brief has concrete legal and custodial consequences.
For law firms evaluating AI tools, the economics now include regulatory risk. A single compliance failure can trigger disciplinary review, fines, malpractice claims, and reputation damage that no efficiency gain offsets. For AI vendors targeting professional services, the lesson is blunt: models marketed as "general-purpose" cannot substitute for domain-specific validation. Enterprises deploying large language models in regulated industries will demand audit trails, confidence scoring, and explicit hallucination detection—features that add cost and complexity.
