New Mexico Lawyer Fined After ChatGPT Invented Witnesses in a Murder Appeal Brief

Key points

• The New Mexico Supreme Court fined attorney Stephen Aarons in 2025 and held him in contempt after he submitted an AI-generated brief containing invented evidence.
• The brief, filed during a murder conviction appeal, included fabricated police testimony and witnesses that ChatGPT made up.
• Aarons told the court he used ChatGPT to produce what he called a "bulletproof summary" of the case.
• OpenAI's ChatGPT is a text-generating AI chatbot that writes convincing prose but sometimes invents facts entirely.
• The court found Aarons hadn't checked whether anything the AI wrote was true.

A lawyer trying to free a client convicted of murder submitted a document to New Mexico's highest court containing witnesses who never existed and police testimony that was never given. All of it came from ChatGPT.

The New Mexico Supreme Court caught the fabrications. On Wednesday it fined attorney Stephen Aarons and held him in contempt, a formal finding that he'd defied the court's standards.

What exactly went wrong?

Aarons used ChatGPT to draft what he called a "bulletproof summary" of the case. ChatGPT writes fluently, but it sometimes invents names, quotes and events that sound real but aren't. Lawyers call this "hallucination," the term for when an AI confidently produces false information.

Aarons submitted the brief without verifying whether the people, testimony or facts it described actually existed. In a murder appeal, where a real person's freedom depends on every word, that omission had serious consequences.

The pattern isn't new to courts. A federal judge in New York dealt with almost the same problem in 2023, when lawyers submitted ChatGPT-generated citations to cases that had never existed. Our earlier report on AI-generated fake submissions flooding Australia's parliamentary inquiry system showed the same failure mode hitting government processes: invented sources attributed to real academics, built into official records.

What does this mean for anyone going through the legal system?

This story's a practical warning if you're involved in any court case. AI tools can help a lawyer draft documents faster, but they can't be trusted to produce accurate facts on their own. A lawyer who skips verification is putting the client's case and their own career at risk.

The responsibility sits entirely with the human professional who files the document. Courts are now enforcing that with contempt findings.

What matters most here isn't the fine. It's that courts are actively hunting for AI-generated errors now, and that scrutiny will intensify. Any lawyer treating these tools as a shortcut rather than a first draft is gambling with a client's future. Aarons' case is the clearest proof yet that the bet doesn't always pay.