By Emma Roth
Published September 11, 2026, 8:44 PM UTC
Main Facts
The New Mexico Supreme Court has issued a severe rebuke and financial penalty to an attorney who relied blindly on artificial intelligence to draft a legal brief for a high-stakes murder appeal. In a filing released on Wednesday, the state’s high court held attorney Stephen Aarons in contempt of court and slapped him with a $5,000 fine.
The disciplinary action stems from a deeply flawed legal brief submitted by Aarons on behalf of a client convicted of murder. Rather than relying on trial transcripts and authenticated legal precedent, Aarons used OpenAI’s ChatGPT to generate the appeal. The resulting document was riddled with catastrophic fabrications, including entirely fictitious witnesses, invented police testimony, and false assertions regarding the shooter’s physical appearance and clothing.
During an August hearing addressing the brief, Justice C. Shannon Bacon sharply questioned Aarons regarding his technological naivety. When Aarons admitted he had used ChatGPT in hopes of creating a "bulletproof summary" of the trial without understanding the inherent risks of generative AI, Justice Bacon retorted:
"Counsel, do you watch the news? Do you listen to the radio? Do you read anything about what’s going on in the world? Because the problem with lawyers relying on AI hallucinations is an above-the-fold story every single day."
The New Mexico ruling is the latest in a mounting wave of judicial penalties levied against legal professionals who fail to exercise basic professional skepticism when employing artificial intelligence tools.
Chronology of the Case
To understand how a murder appeal descended into digital fabrication, it is necessary to trace the timeline of events leading up to the New Mexico Supreme Court’s September ruling:
- The Trial and Conviction: An individual is convicted of murder in a New Mexico trial court, prompting the defense to launch an appeal.
- The Drafting Phase (Early 2026): Attorney Stephen Aarons takes on the appeal. Seeking to streamline the arduous process of synthesizing trial records, Aarons turns to ChatGPT, asking the large language model to produce a summary of the trial proceedings.
- The Brief Submission: Believing he had secured a "bulletproof summary," Aarons incorporates the AI-generated narrative into an official court brief. The document includes detailed accounts from witnesses who never took the stand and fabricates specific testimonies from law enforcement officers.
- Discovery and Scrutiny (August 2026): The New Mexico Supreme Court reviews the brief and identifies glaring discrepancies, non-existent testimonies, and corrupted factual claims. The court summons Aarons to explain the origins of the brief.
- The August Hearing: During oral arguments, Aarons admits to using ChatGPT. Justice C. Shannon Bacon delivers her blistering critique, questioning his awareness of widespread industry warnings regarding AI hallucinations.
- The Ruling (September 11, 2026): The New Mexico Supreme Court officially releases its filing holding Aarons in criminal contempt, assessing a $5,000 fine, and formally condemning his failure to verify the factual and legal accuracy of his filing.
Supporting Data and Industry Context
The New Mexico incident is far from an isolated anomaly; it highlights a systemic vulnerability within the modern legal profession. As generative AI tools become more integrated into daily workflows, the frequency of fabricated legal submissions—commonly referred to in tech and legal circles as "hallucinations"—has escalated sharply.
According to recent data from the American Bar Association (ABA) and industry legal reports, an increasing number of attorneys are experimenting with and deploying large language models for legal research, contract review, and document drafting. However, a significant portion of the legal workforce remains untrained in recognizing the operational limits of these models.
Generative AI models are statistical prediction engines designed to generate human-sounding text based on patterns, not factual databases. When an AI model lacks a specific piece of information, it often invents—or "hallucinates"—plausible-sounding details, complete with convincing case citations, fake docket numbers, and fictional quotes.
High-profile cases of AI misbehavior in courts have accumulated globally over recent years:
- The Two Law Firms Debacle: Last year, a federal judge heavily penalized two law firms for submitting a brief that contained "numerous false, inaccurate, and misleading legal citations and quotations" entirely conjured up by AI.
- The MyPillow Defense Team: Attorneys representing Mike Lindell and MyPillow faced severe judicial fines after incorporating AI-generated misquotes and completely fabricated legal precedents into their court filings.
- The New Mexico Benchmark: The Stephen Aarons case marks a significant escalation because the AI-generated errors were not merely abstract legal citations, but foundational factual fabrications—inventing witnesses and police testimony in a violent crime appeal.
Official Responses and Statements
Faced with mounting public scrutiny and professional censure, attorney Stephen Aarons issued a public statement addressing his actions and the court’s penalty. Speaking to Reuters, Aarons expressed remorse while attempting to contextualize the error as an oversight:
"I am remorseful but hopeful that the disciplinary board takes into account it was an honest mistake," Aarons stated.
Despite his appeals for leniency, the New Mexico Supreme Court remained firm, emphasizing that the baseline duty of an attorney is to independently verify every assertion—both factual and legal—submitted to an officer of the court. The court’s ruling underscores that outsourcing professional responsibility to a chatbot does not absolve a lawyer of professional liability or ethical breaches.
Legal ethics boards across the United States have increasingly updated their guidance, warning attorneys that blind reliance on generative AI violates rules of professional conduct regarding competence, diligence, and candor toward the tribunal.
Implications for the Legal Profession
The New Mexico Supreme Court’s decision carries profound implications for the intersection of law and artificial intelligence. As generative AI transforms from a novel technological experiment into a standard workplace tool, the legal industry faces a reckoning regarding accountability, ethics, and education.
1. The Death of "Plausible Deniability"
Judges across federal and state jurisdictions are signaling zero tolerance for AI-related errors. The defense that "the computer made a mistake" is no longer viewed as a mitigating factor; rather, it is viewed as an aggravating admission of professional negligence. Attorneys are personally and professionally responsible for every word that appears on their signature block.
2. Mandatory AI Literacy and Training
Law firms and bar associations are being forced to reevaluate continuing legal education (CLE) requirements. Future training programs must incorporate modules on prompt engineering, the mechanics of machine learning hallucinations, and rigorous verification protocols. Software tools specifically designed for legal AI—which cross-reference databases and flag unverified citations—may soon shift from optional assets to mandatory professional safeguards.
3. Client Trust and Public Perception
Incidents like the Stephen Aarons case threaten public confidence in the judicial system. When a murder appeal is compromised by a chatbot inventing fictional police officers and eyewitnesses, the integrity of the adversarial process is called into question. Rebuilding that trust requires the legal community to establish transparent, enforceable guardrails around the use of emerging technologies.
As Justice Bacon’s warning reverberates through the legal community, one message is abundantly clear: while artificial intelligence may assist in the practice of law, it can never replace human judgment, verification, and accountability.
