US lawyer cites fake witnesses in murder case, blames ChatGPT

New Mexico’s highest court has sanctioned a Santa Fe defense lawyer after finding that a brief he submitted in a murder appeal relied on entirely fabricated witness statements generated by ChatGPT. The justices concluded the filing contained false police testimony and other invented material and chided the attorney for not confirming the accuracy of what he filed. The lawyer, Stephen Aarons, told the court he used ChatGPT to summarize trial transcripts and other documents when he agreed to take the appeal. At an August 21 hearing Aarons said he expected the AI to produce a reliable summary; the court said he demonstrated both a lack of remorse and insufficient concern for his client’s interests. The panel imposed a $5,000 fine and said it would refer Aarons to a disciplinary board for further investigation. The appeal involves Oscar Renee Sandoval, who was convicted last year and sentenced to life in prison after pleading not guilty to the murder of the mother of his children. The Supreme Court ordered Aarons to explain how the fabricated passages — which the court noted included fictional descriptions of the shooter’s clothing — made their way into his primary brief. The appeal has since been reassigned, on September 2, to public defender Kim Chavez Cook. Court officials and outside observers placed the decision in a broader pattern of judicial pushback against unvetted AI use. The court noted that dozens of attorneys elsewhere have been sanctioned for filing briefs with AI-generated errors such as made-up citations or misquoted law, and some judges have faced scrutiny for their own use of AI. Aarons told Reuters he regretted the mistake and urged that it be seen as an honest error; OpenAI did not immediately respond to requests for comment.