By: PYMNTS
Published: September 13, 2026
Main Facts
The intersection of artificial intelligence and the legal profession reached a stark and sobering milestone last week when the New Mexico Supreme Court formally held defense attorney Stephen Aarons in direct contempt of court. The disciplinary action stems from a high-stakes murder appeal in which Aarons submitted an appellate brief containing entirely fabricated legal authorities and false testimony attributed to wholly made-up witnesses.
Rather than conducting traditional legal research or independently verifying the outputs provided by the technology, Aarons relied blindly on a generative AI tool. He admitted during subsequent court proceedings that he failed to check the factual claims or legal precedents embedded in the brief before filing the document with the state’s highest court. Furthermore, Aarons neglected to inform his client—an individual facing serious criminal penalties—that the brief submitted on their behalf was riddled with machine-generated errors and legal misrepresentations.
In a unanimous and uncompromising response, the New Mexico Supreme Court filed an order on September 9 referring Aarons to the state’s disciplinary board. The bench sharply criticized the veteran attorney, noting in the order that his conduct demonstrated a profound "lack of remorse and a lack of concern for his client."
The incident highlights a growing, perilous trend across the global legal landscape: the uncritical adoption of generative artificial intelligence by legal practitioners who fail to understand, or choose to ignore, the inherent limitations of large language models (LLMs). As corporate legal departments race to institutionalize automated workflows and law schools grapple with how to preserve fundamental analytical skills, individual practitioners continue to run afoul of professional ethics standards by treating generative AI as an infallible oracle rather than a probabilistic assistant.
Chronology of Events
The unfolding controversy surrounding Stephen Aarons’ legal brief traces a clear timeline from technological deployment to judicial reckoning:
- Early 2025: OpenAI releases its advanced o3 model, fueling a new wave of enterprise and professional integration for generative AI systems, touted for their enhanced reasoning capabilities.
- Mid-2026: Tasked with a complex murder appeal in New Mexico, defense attorney Stephen Aarons turns to an instance of ChatGPT powered by the OpenAI o3 model to draft portions of his appellate filing. Assuming the tool would produce a "bulletproof summary of proceedings," Aarons inputs prompts without subsequently auditing the generated citations, case law, or witness testimonies.
- August 2026: During a formal hearing before the New Mexico Supreme Court, justices scrutinize the brief and question Aarons regarding citations and quotes that do not correspond to any known legal precedents or actual court transcripts. Under questioning, Aarons admits he failed to verify the facts, did not notify his client of the generative AI usage, and assumed the technology was foolproof given its widespread deployment across professional sectors.
- September 9, 2026: The New Mexico Supreme Court formally files a written order holding Stephen Aarons in direct contempt of court. The order highlights his lack of remorse, details his failure to protect his client’s interests, and officially refers his case to the state disciplinary board.
Supporting Data and Industry Context
The disciplinary action against Aarons arrives against a backdrop of rapid, sometimes chaotic, technological transformation within the legal sector. While individual practitioners stumble over basic verification protocols, institutional entities are experiencing a massive surge in generative AI adoption.
Data compiled in mid-2026 by FTI Consulting and Relativity reveals a dramatic shift in how corporate legal departments handle technology. According to their findings, generative AI adoption within corporate legal teams nearly doubled over the course of a single year, jumping from 44% in 2025 to an overwhelming 87% in 2026.
Sophie Ross, Global CEO of FTI Technology, summarized the structural shift in the industry, noting that "generative AI has become a fixture in the majority of legal departments." However, this rapid institutional integration stands in stark contrast to the uneven digital literacy and compliance oversight demonstrated by solo practitioners and small-firm lawyers who may lack internal compliance frameworks, rigorous IT auditing, or mandatory continuing legal education on AI ethics.
At the same time, educational institutions are aggressively pushing back against the uncontrolled reliance on automated writing tools. As reported by the Financial Times, law schools worldwide have begun implementing strict classroom bans on generative AI tools. The objective behind these prohibitions is to ensure that students do not stunt their own academic growth, critical thinking abilities, and foundational legal writing skills—core competencies that prevent the exact type of oversight exhibited in the New Mexico appeal.
Official Responses and Judicial Scrutiny
The reaction from the New Mexico Supreme Court bench during last month’s hearing underscores the judiciary’s mounting exhaustion with attorneys attempting to excuse professional negligence by blaming software errors.
During the proceedings, justices expressed outright disbelief that an active attorney could remain ignorant of the well-documented phenomenon of "AI hallucinations"—instances where language models generate plausible-sounding falsehoods, entirely fabricated case law, or non-existent quotes with absolute confidence.
Justice C. Shannon Bacon delivered a stinging critique of Aarons’ defense during the hearing, stating that she was "really struggling with" his claim of complete unawareness regarding generative AI risks. Justice Bacon pointed out that the dangers of relying on unverified AI information constitute a daily "above-the-fold" news story across mainstream media and legal journals alike.
"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," Justice Bacon remarked from the bench.
Aarons, defending his actions to the court, explained that his misplaced trust stemmed from the ubiquity of artificial intelligence in contemporary professional life. "I assumed that it generated a bulletproof summary of proceedings," Aarons stated, explaining that he believed the tool would inherently maintain accuracy because of how widely adopted AI has become in both the legal and medical sectors. The state supreme court, however, firmly rejected this rationale as a breach of basic professional competence and due diligence.
Broader Implications for the Legal Profession
The sanctioning of Stephen Aarons serves as a cautionary tale with far-reaching implications for attorneys, law firms, and the judiciary at large. As generative AI models become increasingly sophisticated—offering advanced reasoning capabilities that often mask underlying errors with authoritative prose—the legal industry faces a profound cultural and ethical reckoning.
1. The Erosion of Competence and the Duty of Diligence
Under model rules of professional conduct across various jurisdictions, lawyers owe their clients a fundamental duty of competence and diligence. This duty inherently includes verifying the factual and legal accuracy of any document submitted to a court of law. Pinning liability or shifting blame onto a third-party software provider—whether it is a word processor, a database, or an advanced generative AI model—has consistently been rejected by courts. The attorney of record remains ultimately responsible for every word, citation, and argument presented in a brief.
2. The Persistence of "Hallucinations"
Despite continuous fine-tuning by developers, large language models remain probabilistic text predictors rather than deterministic databases. They generate responses based on statistical patterns in language, meaning they can—and frequently do—invent convincing case citations, non-existent judicial rulings, and fabricated testimonies. As long as lawyers treat these tools as autonomous researchers rather than assistive draftsmen, incidents of judicial sanction will likely proliferate.
3. Institutionalization vs. Fragmentation
The stark divide highlighted by recent data—where 87% of corporate legal departments have formal generative AI strategies while independent practitioners navigate the technology in isolation—suggests that small firms and solo practitioners are at a disproportionately high risk of ethical violations. Institutional legal departments typically deploy walled-garden enterprise solutions, robust data governance protocols, and mandatory internal training. Conversely, solo attorneys utilizing consumer-facing chatbots without enterprise safety guardrails are uniquely vulnerable to catastrophic professional errors.
4. Rebuilding Judicial Trust
Judiciaries across the United States are increasingly enacting local rules, standing orders, and certification requirements specifically targeting the use of generative AI in court filings. Attorneys are now frequently required to explicitly disclose whether artificial intelligence was used in drafting pleadings and to certify that every citation has been manually verified against official reporters. Incidents like the one in New Mexico threaten to erode judicial trust in electronically prepared briefs, potentially increasing scrutiny and procedural burdens for all practitioners who utilize modern legal technology responsibly.
Ultimately, the New Mexico Supreme Court’s decisive action against Stephen Aarons sends an unequivocal message to the legal community: while artificial intelligence is rapidly reshaping the mechanics of legal practice, human accountability, professional ethics, and rigorous independent verification remain non-negotiable pillars of the justice system.
