The New Mexico Supreme Court held a ChatGPT-using lawyer in direct contempt of court for submitting a brief with “false testimony from wholly fabricated witnesses,” including fake police testimony and other errors. The state’s top court referred the lawyer to a disciplinary board for further proceedings and concluded that he “demonstrated a lack of remorse and a lack of concern for his client.”

Attorney Stephen Aarons “admitted to the Court that he did not verify the factual claims and legal authority in his AI-generated brief before signing it and filing it with the Court, and that he did not inform his client of this failure or that the brief in chief contained multiple factual and legal misrepresentations,” the state Supreme Court said in an order on Wednesday.

Aarons has been a criminal defense lawyer in New Mexico for over 40 years and was hired by a defendant’s family members to appeal a murder conviction. His now-former client, Oscar Renee Sandoval, was sentenced to life in prison in February 2025 after being convicted of killing Shiereen Al-Jibury, who was his partner and the mother of his children.

In August 2025, Aarons submitted the brief containing fake testimony and other errors. Weeks later, the state of New Mexico filed a motion to strike portions of that brief.

“Respondent admitted to the Court that the brief in chief contained false testimony from wholly fabricated witnesses—Officer Michelle Amarillo, Officer Sanchez, Manal Al-Jibury, and Teresa Marquez,” the court order said.

Lawyer Fed Trial Transcript into ChatGPT

Aarons further admitted submitting “false testimony from Danny Stanton that he received threats,” “false testimony from Linda Stanton about the threats her husband received,” and “false testimony from Mariah Chavez and Teresa Marquez (fabricated witness) regarding the shooter’s clothing and appearance.” The order also noted that Aarons “misrepresented legal authority” in citations to previous cases.

Many lawyers have been caught citing fake cases in briefs or inaccurately describing real cases. While Aarons didn’t cite fake cases, he inaccurately described real ones and cited fake testimony.

Aarons told the state Supreme Court at a hearing on August 21 that he fed a computer-generated transcript of the murder trial and other documents related to the case into ChatGPT, which outputted the fake quotes.

“It’s of little comfort to know that my stupidity is what brings us together this afternoon,” Aarons told the court. He admitted his brief quoted “several witnesses who were never called at trial,” another “witness who was called but the brief got the name wrong,” and that his brief inaccurately described precedents.

Aarons indicated that he used a version of ChatGPT powered by the OpenAI o3 model, released earlier in 2025. “I assumed that it generated a bulletproof summary of proceedings,” he said, explaining that he believed it would be accurate given how widespread AI use is in the legal and medical fields.

Aarons was barred from appearing before the New Mexico Supreme Court pending the outcome of any disciplinary board investigation and proceedings. He was fined $5,000, to be paid to the State Bar of New Mexico Client Protection Fund, with additional penalties possible from the disciplinary proceedings. The state Supreme Court ordered the public defender office to appoint a new lawyer for the defendant, struck all previous briefs from the record, and said the case will proceed in the court’s 2026–27 term.

Justice: “You Buried Your Head in the Sand”

Justices lambasted Aarons during the hearing. They expressed surprise that he did not know AI tools could generate false information and pointed out that attorneys must verify the accuracy of information regardless of its source. Whether a lawyer gets help from an AI tool, a law student, or a fellow attorney, the lawyer signing the brief must attest to its accuracy, they said.

Justice C. Shannon Bacon was particularly withering in her criticism. She told Aarons that there are “at least eight or nine provisions in the code of conduct that you violated” and said she was “really struggling with” his claim to be unaware of AI hallucinations, noting that her 13-year-old nephew and 75-year-old stepmother are both aware of the problem.

“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,” she said. “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.”

Aarons responded that he submitted the brief a year ago and “a lot has come out in the last year.” However, the problem of lawyers using AI to cite made-up cases has been in the news regularly for well over three years.

Aarons added that the cases he cited were not fake, though his brief described them inaccurately. Bacon responded that there is no “material distinction” between inaccurately describing a real case and citing a fake one, as the rules about candor to the court apply “with equal force” either way.

”I Didn’t Know That AI Could Hallucinate Facts”

Aarons provided a statement when contacted by Ars Technica. He said:

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.

Aarons told the court that he began by using Rev.com, a service that provides AI transcripts of audio files, and then loaded that transcript and other materials into ChatGPT. “I believe I did confirm” that the trial transcript was accurate, he said when asked whether he verified it before inputting it. “The problem wasn’t in the transcript. It was when I loaded it and all the other information” into ChatGPT. In addition to the transcript, Aarons said he input “the record proper, the statement of issues, and some of the discovery.”

Aarons said he hoped the court would take steps to prevent other lawyers from making the same mistake, such as issuing a standing order requiring briefs to include a certificate of compliance regarding AI use. Justices treated the suggestion as a distraction from the core problem: his failure to verify the accuracy of the brief.

“Assume with me that you had relied on the work of a first-year lawyer that was working for you and that they had just made stuff up… and you signed it. You’d be in the same exact soup you are right now,” Bacon said. “So the idea that, the suggestion in your briefing that because the court didn’t tell you at the time that you did this, ‘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."

"Your Client Is the One Suffering”

“The other thing that’s completely missing from your response is anything about what this has done to your client,” Bacon continued. “That’s who I’m worried about. You have presented briefing to the court that we cannot rely on, and your client is the one suffering because of this far more than you will ever suffer.”

Chief Justice Julie Vargas similarly told Aarons: “I’m really interested with the approach you’re taking in this hearing. 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. And I really don’t care about the policy concerns right now. I care about what we’re going to do with your client and what we should do in this circumstance based on the behavior that you showed us.”

Justice Michael Vigil said there is nothing wrong with using AI to help write a legal argument as long as the lawyer verifies its accuracy. “It doesn’t matter what the tool is. It doesn’t matter how advanced the AI-generated program is or what improvements they make, whatever,” he said. “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.” The case now returns to the court’s docket with new counsel appointed, leaving Sandoval’s appeal to restart from scratch.