New Mexico's Supreme Court has sanctioned an attorney for filing a murder appeal built on witnesses who never existed — fabricated wholesale by ChatGPT — in one of the starkest examples yet of artificial intelligence "hallucinations" reaching the heart of the American justice system.
According to a report from Reuters, the state's highest court fined Santa Fe lawyer Stephen Aarons $5,000 and held him in contempt for failing to "verify the factual claims and legal authority in his AI-generated brief." The brief, submitted as part of an appeal against his client's murder conviction, "contained false testimony from wholly fabricated witnesses," along with false claims about the shooter's clothing and appearance. For more context on this story, see our ongoing breaking AI news.
What the Court Found
The court's filing did not describe minor inaccuracies. It described an appellate document that conjured people out of thin air: testimony attributed to witnesses who do not exist, and invented police statements about what the shooter wore and how the shooter looked — precisely the kind of factual detail a murder conviction can turn on.
According to The Verge, which reported on the court filing, the sanction was imposed because Aarons failed to "verify the factual claims and legal authority in his AI-generated brief" before submitting it to the court. In other words, the problem was not that a lawyer used an AI tool to help with drafting. It was that the lawyer filed the output without checking a single fabricated claim against the actual trial record.
"Counsel, Do You Watch the News?"
The extraordinary nature of the case surfaced at an August hearing, where Aarons admitted he had used ChatGPT and told the court he had expected the chatbot to produce a "bulletproof summary" of the trial. That expectation drew a pointed response from New Mexico Supreme Court Justice C. Shannon Bacon.
"Counsel, do you watch the news? Do you listen to the radio? Do you read anything about what's going on in the world?" Bacon asked, according to Reuters. "Because the problem with lawyers relying on AI hallucinations is an above-the-fold story every single day."
The justice's rebuke captured the extent to which AI-fabricated citations and testimony have moved from a niche technical problem to a mainstream professional liability — one that judges now assume every practicing attorney should know about.
The Latest in a String of AI Sanctions
The New Mexico case is the latest in a rapidly growing list of courts punishing lawyers for unchecked AI use. Last year, a judge slammed two law firms for submitting a brief containing "numerous false, inaccurate, and misleading legal citations and quotations," while lawyers representing MyPillow founder Mike Lindell were also fined for including AI-generated misquotes and fake citations in a court filing, The Verge noted.
What sets the New Mexico sanction apart is the material that was fabricated. Most earlier cases involved invented legal citations — fake precedents, bogus quotes from real cases. Here, the fabrication went to the facts themselves: witnesses, testimony, and physical evidence descriptions in a criminal appeal. That distinction matters, because fact fabrication in a murder case strikes at the core of what appellate review exists to protect.
The trend line is unmistakable. As The Verge observed, with the number of lawyers using AI tools in their work increasing, so are the instances in which a court discovers citations that are fabricated — or hallucinated — by AI. What was once treated as an embarrassing oversight is increasingly punished as a failure of professional competence, with fines, contempt rulings, and scrutiny from disciplinary authorities.
"I Am Remorseful"
Aarons, for his part, told Reuters he hopes the fallout stops short of professional discipline. "I am remorseful but hopeful that the disciplinary board takes into account it was an honest mistake," he said in a statement.
Whether the disciplinary board treats it as an honest mistake remains to be seen, but the state's highest court has already made its own view clear: a $5,000 fine and a contempt holding for a brief that was never verified against reality.
Why It Matters
The case lands as courts across the country are hardening their positions on AI use in litigation. Judges have increasingly required attorneys to certify that AI-generated filings have been verified, and sanctions have become the visible enforcement mechanism. The message from New Mexico's Supreme Court is likely to travel well beyond its borders: the duty to verify is personal, non-delegable, and does not disappear because a chatbot produced confident-sounding prose.
For the legal profession, the calculation is shifting. AI drafting tools can save hours of work, but a single fabricated witness in a filed brief can now cost a fine, a contempt ruling, and a referral to disciplinary authorities — a trade few practitioners will find attractive. For everyone else, the case is a reminder that the people meant to check AI's outputs are under more pressure than ever to actually check them.
The stakes are highest in criminal cases. Sanctions over fake citations waste judicial time; fabricated evidence in a murder appeal touches the mechanism that decides whether a conviction stands. An appellate court can only weigh claims anchored to the real record — and a brief seeded with phantom witnesses asks judges to rule on events that never happened.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
