New Mexico Supreme Court fines lawyer $5,000 after AI-generated fake witnesses appear in murder appeal

Updated September 12, 2026 12:24 am

New Mexico’s Supreme Court has fined attorney Stephen Aarons $5,000 and held him in contempt after his appeal in a murder case included AI-generated witnesses, false testimony and fabricated legal support. The ruling is a stark warning that courts are no longer treating ChatGPT mistakes as harmless shortcuts when they distort the record in serious criminal cases.

The sanction matters well beyond one appeal: it signals that judges are increasingly prepared to punish lawyers who submit AI-produced material without verifying it, especially when the errors involve invented evidence and citations that can mislead a court.

What the court said happened

The state’s highest court said Aarons failed to check the accuracy of a brief that contained material generated with the help of ChatGPT. According to the court filing reported by Reuters, the submission included false claims attributed to nonexistent witnesses, as well as inaccurate descriptions of the shooter’s clothing and appearance.

The court concluded that the brief did not merely contain sloppy citation errors. It said the filing relied on fabricated facts and legal authority that should have been checked before submission, particularly in an appeal tied to a murder conviction.

Aarons told Reuters he was remorseful and hoped disciplinary authorities would treat the matter as an honest mistake.

The court’s response, however, suggests it viewed the lapse as much more serious than a routine mistake. By issuing a monetary penalty and a contempt finding, the justices underscored their view that lawyers cannot outsource verification to AI tools and then blame the software when the result is wrong.

Why did the judge’s reaction draw attention?

The reaction drew attention because the exchange captured a wider concern spreading through the legal profession: lawyers increasingly use generative AI, but courts are finding that many still do not understand how easily these systems can invent facts, names and legal authorities.

At an August hearing, Justice C. Shannon Bacon sharply questioned how a lawyer could be unaware of the problem. According to Reuters, Bacon asked Aarons whether he followed the news or paid attention to the public discussion around AI hallucinations in law. Her point was blunt: the danger is now common knowledge, and the responsibility to verify remains with the attorney.

Justice C. Shannon Bacon said the issue of lawyers relying on AI hallucinations is widely reported and should not have surprised counsel.

The exchange reflects a growing judicial impatience with lawyers who treat chatbots like reliable research assistants. In the court’s view, that is no longer a credible excuse once false citations or made-up witnesses reach an appellate filing.

How did ChatGPT become part of a murder appeal?

ChatGPT became part of the case when Aarons used it while preparing the appellate brief and, by his own account, expected it to produce a strong summary of the trial. He told the court he believed the tool would give him a “bulletproof” version of the facts.

That expectation captures the main risk of generative AI in legal work: the output can sound polished and confident even when it is wrong. In a high-stakes criminal appeal, that can create especially dangerous consequences because the court depends on counsel to present accurate records and trustworthy citations.

The New Mexico ruling shows how quickly a convenience tool can become a liability. If a lawyer uses AI to speed up drafting but does not independently confirm every fact, the brief can blend real material with falsehoods in a way that is difficult to detect at first glance.

What are AI hallucinations in legal briefs?

AI hallucinations are outputs in which a model produces invented or misleading information that appears plausible. In legal briefs, that can mean fake cases, incorrect quotations, nonexistent witnesses or false summaries of the evidence.

Because legal writing relies on precision, even a small hallucination can undermine an entire filing. A single fabricated citation can damage credibility, waste judicial resources and expose the lawyer to sanctions or disciplinary review.

How unusual is this kind of sanction?

The fine is part of a broader pattern of courts punishing AI misuse in legal filings. It is not the first time judges have responded harshly when attorneys submitted documents containing false citations or invented language generated by AI tools.

Last year, a judge criticized two law firms for filing a brief containing numerous false and misleading citations and quotations. In another widely reported case, lawyers representing MyPillow founder Mike Lindell were fined after their brief included AI-generated misquotes and citations that did not exist.

Those examples matter because they show the issue is no longer hypothetical. Courts are now seeing real filings with AI-created errors, and judges are making clear that “I asked a chatbot” is not a defense.

Why this case matters for the legal profession

This case matters because it highlights a simple truth: AI can help lawyers draft faster, but it cannot be trusted as a source of legal facts or authority without human review. The New Mexico Supreme Court’s action sends that message in especially forceful terms.

As more attorneys experiment with generative AI, the risk is not limited to bad research. A lawyer may unintentionally file a brief that appears professional but contains fake testimony, distorted details or invented precedent. That can mislead judges, harm clients and trigger professional discipline.

The ruling also reinforces a central ethical rule of legal practice: responsibility cannot be delegated to software. Even if an AI model contributed to the error, the filing is still the lawyer’s duty to verify.

What lawyers should take from the ruling

Lawyers should treat AI as a drafting aid, not a source of truth. Before any filing leaves the office, counsel should independently confirm every quote, citation, factual statement and record reference.

• Check all citations against authoritative legal databases.
• Verify factual assertions against the trial record.
• Review AI-generated summaries for invented details.
• Use internal supervision and quality control before filing.

For firms, the case is also a reminder to train staff on the limits of generative AI. A policy that merely permits AI use is not enough unless it also requires verification and review procedures.

What did Aarons say after the ruling?

Aarons said he was sorry and hoped the disciplinary process would recognize the filing as an honest error, according to Reuters. That response may help explain intent, but it does not eliminate the consequences of putting false material before a court.

The distinction matters because disciplinary bodies often weigh both misconduct and harm. Even where there is no sign of deliberate deception, an attorney can still be sanctioned if the filing shows a serious failure to exercise basic professional care.

In this case, the court appears to have decided that the combination of fabricated witnesses, inaccurate testimony and unsupported legal claims crossed that line.

Timeline: How the case unfolded

Why judges are growing more aggressive about AI errors

Judges are becoming more aggressive because AI-related filing mistakes waste time, threaten fairness and can compromise the integrity of the judicial process. When a court must sort fact from invention, everyone pays the price: the bench, opposing counsel, clerks and the public.

There is also a practical reason for the tougher stance. If judges respond lightly, lawyers may conclude that AI-generated errors carry little risk. Strong sanctions create a deterrent and set a clear expectation that filing standards have not changed simply because the drafting tool has.

The New Mexico case fits that trend. The court did not frame the issue as a novelty or a technical hiccup. It treated the problem as a serious professional failure that deserved a public penalty.

How common are AI citation problems in court filings?

AI citation problems are becoming common enough that many judges now recognize them immediately. The legal system has already seen multiple incidents in which attorneys filed documents with cases, quotations or authorities that did not exist.

The pattern is especially visible in jurisdictions where lawyers have adopted ChatGPT and similar tools for research and first-draft generation. Because these systems are built to produce plausible language rather than guaranteed accuracy, they can mimic legal writing while quietly introducing major errors.

That combination makes AI particularly dangerous in law. A brief may look polished enough to pass an initial review, yet still contain invented details that undermine the entire case once challenged.

What this means for AI use beyond the courtroom

The ruling is a warning for any profession that relies on accuracy, documentation and traceability. If lawyers can be sanctioned for failing to verify AI output, the same logic may apply in compliance, journalism, finance and other fields where factual mistakes can have serious consequences.

The core lesson is not that generative AI cannot be used. It is that it must be used with skepticism, supervision and systematic fact-checking. The more authoritative the setting, the less room there is for blind trust.

For courts in particular, the case may accelerate the push toward explicit AI-use rules, staff training and filing checklists designed to catch hallucinated material before it reaches a judge.

Bottom line

The New Mexico Supreme Court’s $5,000 sanction against Stephen Aarons is one of the clearest recent signs that courts expect lawyers to verify every AI-assisted claim before filing. In a murder appeal, where accuracy is essential, fabricated witnesses and false testimony were enough to trigger contempt and a public warning to the profession.

As more attorneys experiment with generative AI, the message from New Mexico is straightforward: if a chatbot helps draft the brief, the lawyer still owns every word.

Key facts at a glance