The New Mexico Supreme Court has sanctioned criminal defense attorney Stephen Aarons, fining him $5,000 and holding him in contempt for filing an appellate brief that contained fabricated witnesses and police testimony generated by ChatGPT. The court also removed Aarons from his client's appeal of a life sentence for murder, according to a report from The Verge.
The sanctions, filed on September 11, 2026, stem from Aarons's failure to "verify the factual claims and legal authority in his AI-generated brief." The court found the document "contained false testimony from wholly fabricated witnesses," as well as invented details about the shooter's clothing and appearance.
During an August hearing, Justice C. Shannon Bacon directly questioned Aarons on his awareness of the well-documented risks of using generative AI in legal work. "Counsel, do you watch the news? Do you listen to the radio? Do you read anything about what’s going on in the world?" Justice Bacon asked, according to Reuters. "Because the problem with lawyers relying on AI hallucinations is an above-the-fold story every single day." Aarons admitted to using ChatGPT and said he believed it would produce a "bulletproof summary" of the trial.
This case marks a significant escalation in judicial response to the misuse of AI in legal proceedings. While lawyers in other jurisdictions have faced fines and professional discipline for submitting AI-generated briefs containing fictitious case citations—a phenomenon known as "hallucination"—the New Mexico matter involves the fabrication of core factual evidence in a high-stakes criminal appeal. Submitting false witness testimony and police statements in a murder case represents a more severe breach of ethical and procedural rules than citing non-existent legal precedents.
The incident is not an isolated one within New Mexico's court system. According to the source material, state courts have detected AI hallucinations in filings for at least seven separate lawsuits since 2023. This pattern indicates a growing, systemic problem as generative AI tools become more accessible and are misapplied by professionals who either do not understand their limitations or fail to perform the necessary oversight.
The sanctions against Aarons serve as a stark warning to the legal profession about the non-negotiable duty of verification. Generative AI models like ChatGPT are designed to produce statistically plausible text, not factual accuracy. They have no ability to discern truth, access real court records, or interview actual witnesses. Lawyers bear ultimate responsibility for the contents of any document they submit to a court, regardless of the tools used in its drafting. The $5,000 fine and contempt ruling underscore that "the computer made a mistake" is not a valid defense against charges of submitting false evidence.
This ruling arrives as courts across the United States are scrambling to establish formal rules and guidelines for the use of AI in litigation. Many federal and state courts have begun implementing local rules that mandate attorneys attest to the fact that they have personally reviewed any AI-generated content for accuracy or disclose its use. The New Mexico Supreme Court's decisive action in a case involving fabricated evidence in a capital-level appeal will likely be cited as a precedent, pushing more jurisdictions to adopt clear, enforceable standards.
The broader implication for the AI industry is increased scrutiny on the real-world harm caused by model hallucinations. While developers often discuss hallucinations as a technical limitation to be solved, this case illustrates their concrete consequences: undermining judicial integrity, violating a defendant's right to competent counsel, and wasting court resources. As AI becomes embedded in high-stakes professional fields like law, medicine, and finance, the pressure on companies to mitigate hallucinations and on professionals to implement rigorous human oversight will only intensify. The New Mexico ruling is a clear signal that the legal system will not tolerate the use of AI as a shortcut around fundamental duties of diligence and truthfulness.
