The rapid integration of generative artificial intelligence into legal practice has crossed a threshold from speculative warning to systemic crisis. In a landmark ruling that reverberates across the legal technology landscape, the New Mexico Supreme Court has held veteran criminal defense attorney Stephen Aarons in direct contempt of court. Aarons submitted an appellate brief on behalf of a convicted murderer containing a cascade of egregious errors, most notably "false testimony from wholly fabricated witnesses" and completely ghost-written legal citations generated by OpenAI’s o3 model.
The state’s highest court did not mince words. In addition to slapping Aarons with a $5,000 fine payable to the State Bar of New Mexico Client Protection Fund, the justices indefinitely barred him from appearing before the court pending an exhaustive investigation by a disciplinary board. The defective brief was unceremoniously struck from the court’s record, and the public defender’s office was ordered to appoint new counsel for the appellant, Oscar Renee Sandoval, whose path to justice has been severely derailed by his own lawyer’s technical negligence.
This high-profile disciplinary action is far from an isolated incident. Across federal and state courts nationwide, judges are increasingly forced to grapple with attorneys submitting filings laced with "hallucinations"—fabricated case law, phantom precedents, and now, entirely invented eyewitness testimonies. Yet, Aarons’ case stands out not merely for the depth of the deception, but for the stunning defense offered by a practitioner with four decades of experience: “I didn’t know that AI could hallucinate facts.”
The ensuing courtroom confrontation between Aarons and the New Mexico Supreme Court justices offers a sobering case study in professional accountability, the hazards of uncritical technology adoption, and the absolute non-negotiability of the duty of candor.
Detailed Chronology of a Legal Trainwreck
The sequence of events leading to the New Mexico Supreme Court’s disciplinary order exposes a breakdown of professional oversight stretching from the spring of 2025 to the harsh reckonings of late summer.
The Conviction and the Appeal
In February 2025, Oscar Renee Sandoval was sentenced to life in prison plus four and a half years following a high-profile first-degree murder conviction. Sandoval had been found guilty of killing Shiereen Al-Jibury, his romantic partner and the mother of his children, in Mesilla, New Mexico. Seeking to overturn the conviction, members of Sandoval’s family retained Stephen Aarons—a criminal defense lawyer who had practiced in New Mexico for over forty years—to handle the appeal.
The Insertion of the Machine
Tasked with drafting the appellant’s brief, Aarons turned to modern technological shortcuts. According to his disclosures to the court, Aarons first utilized Rev.com, an AI-driven transcription service, to convert the audio files of the lengthy murder trial into text.
Rather than reviewing the transcript against the physical reality of the trial record, Aarons fed the computer-generated transcript, along with case discovery, the statement of issues, and the record proper, directly into an advanced iteration of ChatGPT powered by OpenAI’s o3 model. Believing he had secured a "bulletproof summary of proceedings," Aarons compiled the machine’s output into a formal brief and signed his name to it. He filed the document with the New Mexico Supreme Court in August 2025.
The Discovery of Fabrications
Weeks after submission, attorneys for the State of New Mexico filed a motion to strike portions of the brief, having uncovered glaring discrepancies between the appellate claims and the actual trial court record.
When the state scrutinized Aarons’ citations and narrative, they discovered that the brief relied on testimony from people who had never set foot in a courtroom. The document cited "false testimony from wholly fabricated witnesses," including:
Officer Michelle Amarillo
Officer Sanchez
Manal Al-Jibury
Teresa Marquez
Furthermore, the brief cited fabricated accounts of threats received by trial participants, attributing fictitious testimony to real individuals such as Danny Stanton and Linda Stanton, and misstated witness descriptions regarding the shooter’s clothing and appearance via made-up personas like Mariah Chavez. Alongside these fabricated testimonies, Aarons’ brief misrepresented the legal holdings of actual precedent cases.
The August 21 Hearing and Contempt Ruling
Faced with motion papers outlining these bizarre fabrications, the New Mexico Supreme Court summoned Aarons for a disciplinary hearing on August 21, 2025.
During the hearing, Aarons offered a candid but flabbergasting confession. "It’s of little comfort to know that my stupidity is what brings us together this afternoon," he told the justices. He admitted that he had never informed his client—who speaks Spanish and was left entirely in the dark—that the brief submitted on his behalf was structurally and factually fraudulent.
Following the hearing, the court issued a definitive order on Wednesday, ruling Aarons in direct contempt. The penalties were immediate and severe:
Suspension: Barred from appearing before the New Mexico Supreme Court pending disciplinary board findings.
Financial Penalty: Fined $5,000, payable to the State Bar Client Protection Fund.
Remedial Action: All previous briefs filed by Aarons were stricken from the record.
Client Relief: The public defender’s office was directed to appoint competent new counsel for Sandoval to ensure his appeal proceeds unhindered during the 2026–27 term.
Supporting Context & Metrics: The Broader Landscape of AI Hallucinations in Law
Aarons’ defense—that he was entirely unaware generative AI could invent facts—runs counter to a mounting body of case law and professional warnings that have accumulated over the past three years.
A Precedent of Professional Blunders
The phenomenon of lawyers falling victim to AI hallucinations first burst into mainstream legal consciousness in May 2023, when New York attorneys Steven Schwartz and Peter LoDuca made international headlines for submitting a brief containing six entirely fake judicial opinions generated by ChatGPT. Since then, courts across federal and state jurisdictions have levied sanctions, monetary fines, and mandatory remedial education against attorneys who fail to verify AI-assisted work product.
Case / Incident Year
Jurisdiction
Core Infraction
Disciplinary Action Taken
May 2023
U.S. District Court (S.D.N.Y.)
Citing 6 fake cases generated by ChatGPT.
$5,000 fines, mandatory notifications to judges whose names were forged.
Late 2024
Various Federal Circuits
Citing nonexistent appellate holdings and mischaracterizing real statutes.
Show-cause orders, professional reprimands, financial sanctions.
August 2025
New Mexico Supreme Court
Submitting fake witness testimony, fabricated police accounts, and altered precedents via OpenAI o3.
Direct contempt, $5,000 fine, suspension pending board review, brief stricken.
The Nature of "Hallucinations" in Large Language Models
Generative AI models are fundamentally probabilistic text-prediction engines, not databases of verified facts. Trained to predict the most statistically probable next word in a sequence, these models synthesize coherent narratives even when factual anchors are missing. When fed unstructured, lengthy transcripts and asked to summarize or argue points of law, models like OpenAI’s o3 can seamlessly "invent" dialogue, reconcile contradictory testimonies into fictional consensus, and craft plausible-sounding case citations that do not exist in any reporter system.
Legal ethics scholars point out that while technology companies routinely issue safety and capability disclaimers regarding hallucinations, the professional burden of truth rests squarely on the attorney of record. Under standard rules of professional conduct, an AI tool is legally indistinguishable from a junior associate, a paralegal, or an outsourced typing service: the signer assumes absolute liability for every word.
Official Statements and Courtroom Confrontations
The transcript of the August 21 hearing before the New Mexico Supreme Court captures a rare moment of judicial incredulity and direct critique from the bench. The justices systematically dismantled Aarons’ technical excuses, focusing heavily on the betrayal of his duty to his incarcerated client.
Justice C. Shannon Bacon: "You Buried Your Head in the Sand"
Justice C. Shannon Bacon delivered some of the most scathing remarks of the session, expressing utter disbelief at Aarons’ claim of ignorance regarding AI limitations.
"So 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," Justice Bacon demanded. "So either you buried your head in the sand—and that’s a choice to do that, an intentional choice to be uninformed—or you took a gamble, and neither of those are consistent with the code of conduct."
Bacon noted that she was "really struggling" with Aarons’ plea of ignorance, observing dryly that both her 13-year-old nephew and 75-year-old stepmother were well aware that AI tools invent information.
When Aarons attempted to suggest that the court should implement a standing certification order requiring attorneys to disclose AI usage—effectively arguing that courts should establish new guardrails to protect lawyers from themselves—Bacon flatly rejected the premise. Comparing his reliance on ChatGPT to utilizing the unverified work of a negligent first-year associate, she stated:
"Assume with me that you had relied on the work of a first-year lawyer… and that they had just made stuff up… and you signed it. You’d be in the exact exact soup you are right now. So the suggestion… that because the court didn’t tell you at the time… ‘be careful,’ it somehow relieves you of obligation, falls on absolute deaf ears because the rules of professional conduct already tell you what your obligation is."
Chief Justice Julie Vargas: Prioritizing the Client
Chief Justice Julie Vargas similarly brushed aside Aarons’ macro-level policy suggestions regarding the future regulation of AI, refocusing the court’s attention on the immediate human cost of his actions.
"I’m really interested with the approach you’re taking in this hearing," Chief Justice Vargas said. "You seem to be telling us all the policy that we’ve been thinking about for years about what to do with AI, but you’re not talking about how to address the situation that’s in front of us, which has significant impacts on a criminal defendant who is in custody, who’s going to stay in custody until we resolve this matter."
Justice Michael Vigil: The Tool is Irrelevant
Justice Michael Vigil emphasized that the specific technology utilized is entirely secondary to the fundamental failure of verification.
"It doesn’t matter what the tool is. It doesn’t matter how advanced the AI-generated program is or what improvements they make… It doesn’t matter whether you use a C-student lawyer or an A-student lawyer [if] you didn’t check their work before you filed the brief. That’s the issue."
Stephen Aarons’ Defense and Public Statement
In his statements to the court and in subsequent communications with legal media outlets, Aarons expressed remorse while maintaining that his error stemmed from a lack of technical literacy rather than malicious intent.
In a statement provided following the ruling, Aarons noted:
"In March 2025 I agreed to handle an appeal and used ChatGPT to summarize the trial proceedings. I wrote the brief but the table of contents and the summary contained numerous errors. At the time, I didn’t know that AI could hallucinate facts not only in my brief but in pleadings submitted by other attorneys. I am glad the court threw out my defective pleading and ordered the public defender to write a new brief on behalf of my former client. As for myself, I am remorseful but hopeful that the disciplinary board takes into account it was an honest mistake. It is a lesson learned for all professionals who rely upon this powerful but sometimes unstable technology."
However, revelations during the hearing compounded his professional exposure. Aarons admitted that he had not informed his client, Oscar Renee Sandoval, that the brief submitted to secure his freedom was built on AI-generated falsehoods. When asked by Justice Bacon if he had explained the crisis to Sandoval, Aarons replied: "I have not. He’s a Spanish speaker and I wanted to see what happened today." Furthermore, he acknowledged that while he informed the defendant’s family members that there was "a problem with the brief," he purposefully withheld any mention of ChatGPT.
Future Outlook: Implications for the Legal Profession
The disciplinary action against Stephen Aarons marks a critical inflection point for the legal industry, signaling that courts are rapidly losing patience with technical illiteracy as a defense against professional misconduct.
1. The Death of the "Ignorance Defense"
For the past several years, early adopters and technological laggards alike could occasionally plead novelty when caught submitting AI-hallucinated citations. The New Mexico Supreme Court’s ruling effectively closes this window. With generative AI deeply embedded in public consciousness and professional discourse, claiming total unawareness of algorithmic hallucinations is no longer viewed as an unfortunate oversight, but as an intentional dereliction of the duty of competence.
2. Mandatory Disclosure and Local Rules
As courts reject retroactive excuses, legal jurisdictions nationwide are accelerating the adoption of formal standing orders. While the New Mexico justices dismissed Aarons’ suggestion for a specialized AI certificate as a distraction from basic competence, numerous federal judges and state bar associations have already instituted mandatory certification requirements. Attorneys are increasingly required to explicitly state whether generative AI tools were used in drafting briefs and to certify that every citation and factual assertion has been manually verified against the primary record.
3. The Human Toll of Technological Shortcuts
Beyond regulatory penalties and public embarrassment, the Aarons case underscores the profound human cost of algorithmic shortcuts in high-stakes litigation. While lawyers face fines, suspensions, and reputational damage, criminal defendants—individuals whose liberty, appeals, and constitutional rights hang in the balance—suffer the direct consequences of compromised advocacy.
As artificial intelligence tools grow more sophisticated, seamlessly blending fiction with hyper-realistic legal prose, the legal profession faces an unyielding imperative: technology may accelerate the draft, but human diligence remains the ultimate guardian of justice. For Stephen Aarons and the New Mexico Supreme Court, that lesson was learned far too late.
