A New Mexico lawyer has been sanctioned after a court found that his appeal brief contained invented witnesses and fabricated police testimony that appeared to have been generated by ChatGPT.
The New Mexico Supreme Court fined attorney Stephen Aarons $5,000 and held him in contempt for submitting the material without properly checking whether the information was genuine. The court also referred him to the state’s attorney disciplinary authorities for further review.
According to the court, the filing included testimony attributed to witnesses who did not exist, along with fictional details about the murder investigation. Among the fabricated claims was an assertion that the shooter had been dressed in dark pants and a white shirt.
Aarons acknowledged using ChatGPT while preparing the appeal. He said he had taken on the case last year and used the AI system to summarize a computer-generated transcript and other trial materials.
He told the court that he had expected the technology to produce what he described as a reliable summary and said he did not appreciate how extensively the system could invent information.
In a statement, Aarons called the episode an honest mistake and expressed remorse, while saying he hoped the disciplinary authorities would consider the circumstances. He described the incident as a warning for professionals using increasingly powerful AI tools.
The Supreme Court, however, was sharply critical of his failure to verify the material before placing it in a legal filing. The justices said the dangers of AI-generated false information have become widely known and questioned how an experienced lawyer could remain unaware of them.
The case involves Oscar Renee Sandoval, who was convicted of murdering the mother of his children and sentenced to life in prison after pleading not guilty. His appeal remains pending and has now been assigned to New Mexico public defender Kim Chavez Cook.
The incident highlights a growing problem for courts as lawyers increasingly turn to generative AI for research, drafting and document analysis. Judges across the United States have already imposed sanctions in cases where attorneys submitted AI-generated legal citations, quotations or arguments that turned out to be false.
Aarons’ case stands out because the AI-generated material did not merely contain incorrect legal references. It created supposed witnesses and testimony in a criminal case.
The New Mexico Supreme Court had previously ordered Aarons to explain how the fictional material made its way into his principal appellate brief. During an August hearing, he acknowledged feeding the transcript and other case documents into ChatGPT and relying on its output.
The justices made clear that using AI does not remove a lawyer’s responsibility to verify what ultimately reaches the court.
The sanction leaves the underlying murder appeal unresolved, while Aarons now faces a separate disciplinary inquiry over his handling of the case.
