As a lawyer, the last place you’d want to slip up is in the courtroom when your defendant’s fate is at stake. Unfortunately for New Mexico attorney Stephen Aarons, the courtroom was exactly where he messed things up for his career. And it was much more than an accidental slip of the tongue; he had ChatGPT create a list of “wholly fabricated witnesses” while attempting to do a murder appeal for his defendant.
Aarons thought ChatGPT could save him some time and give him an accurate summary of his case to share with the New Mexico Supreme Court. But according to Reuters, he hadn’t taken into account the risks of an LLM “hallucinating” evidence, essentially making things up. Even the justices couldn’t believe a lawyer going into a murder case in 2026 wouldn’t know about the mistakes AI can generate when it’s attempting to compile information.
The lawyer was now guilty of submitting lies to the court and even admitted to saying he didn’t check in with his defendant beforehand to ensure anything was accurate. His excuse is that his defendant is currently in prison and only speaks Spanish, but the judge wasn’t having it, accusing Aarons of being intentionally misinformed and taking a gamble with a case that has someone’s life on the line.
All this could have been avoided had Aarons done a quick glance over ChatGPT’s summary and list of witnesses before taking it to court. However, his refusal to do a basic double-check of the information says a lot about his lack of competence in court, and the Supreme Court is punishing him for it.
The New Mexico attorney was fined $5,000, removed from the case, and referred to the state’s attorney disciplinary board, all because he tried to take a careless shortcut with ChatGPT. It wasn’t a mistake to use an AI tool to make his job easier; it was a mistake to trust his entire career to it without checking facts himself.
