Lawyer fined for AI-generated false testimony in murder appeal
New Mexico’s Supreme Court fined defense attorney Stephen Aarons $5,000 and held him in contempt after his appellate brief contained fabricated witness statements and false details about the shooter’s appearance that were generated by ChatGPT.
Short answer: New Mexico’s Supreme Court fined defense attorney Stephen Aarons $5,000 and held him in contempt after his appellate brief contained fabricated witness statements and false details about the shooter’s appearance that were generated by ChatGPT.
Lawyer fined for AI-generated false testimony in murder appeal
New Mexico’s Supreme Court has penalized a defense attorney for submitting false information that was produced by an artificial intelligence tool. The court ordered Stephen Aarons to pay a $5,000 fine and found him in contempt after discovering that his appellate brief contained fabricated witness statements and invented details about the shooter’s appearance. The filing, which was submitted on a Wednesday, said the brief included testimony from witnesses that do not exist and false descriptions of the clothing and features of the person accused of the shooting.
During an August hearing, Justice C. Shannon Bacon confronted Aarons about his reliance on the AI system. She asked whether he follows news outlets, listens to the radio, or reads any coverage of current events, pointing out that problems with AI hallucinations appear regularly in mainstream coverage. Aarons acknowledged that he had used ChatGPT to help draft the brief, believing it would produce a reliable summary of the trial record. He said he did not realize the model could invent facts that have no basis in reality.
The court’s order emphasized that attorneys have a duty to verify every factual claim and legal citation before submitting documents to the tribunal. It noted that the brief’s falsehoods undermined the integrity of the judicial process and could have misled the justices reviewing the appeal. The sanction serves as a reminder that technological shortcuts do not excuse professional responsibility.
Attorney expresses regret and hopes mistake seen as honest error
Aarons later issued a statement through Reuters expressing regret but hoping that the disciplinary board will consider the mistake an honest error rather than intentional misconduct. He said he remains committed to learning how to use AI tools safely and will take steps to prevent similar issues in the future.
The case is part of a broader pattern in which legal professionals are increasingly turning to generative AI for research and drafting, only to encounter inaccurate outputs. In the previous year, a federal judge reprimanded two law firms after discovering that their briefs contained numerous false citations and quotations that were invented by AI. Separately, attorneys representing the founder of MyPillow were also sanctioned for inserting AI-generated misquotes and fabricated legal references into a court filing.
These incidents have prompted bar associations and courts to issue guidance on the appropriate use of AI in legal work. Many jurisdictions now require lawyers to confirm that any content produced by an AI system is accurate and to disclose when such technology has been employed. Critics warn that without proper oversight, the convenience of AI could erode trust in the legal system, while proponents argue that the technology can improve efficiency when used with caution.
For developers and users of AI, the episode highlights the importance of building models that minimize hallucination and of providing clear warnings about their limitations. It also underscores the need for education programs that teach professionals how to verify AI-generated information before relying on it in high-stakes settings such as litigation. As AI becomes more embedded in everyday workflows, the responsibility to ensure its outputs are trustworthy falls on both the creators of the technology and the people who deploy it.
Now count words. Let's approximate.
I'll count manually roughly.
Details of AI-generated false testimony in the appellate brief
New(1) Mexico’s2 Supreme3 Court4 has5 penalized6 a7 defense8 attorney9 for10 submitting11 false12 information13 that14 was15 produced16 by17 an18 artificial19 intelligence20 tool.21 The22 court23 ordered24 Stephen25 Aarons26 to27 pay28 a29 $5,00030 fine31 and32 found33 him34 in35 contempt36 after37 discovering38 that39 his40 appellate41 brief42 contained43 fabricated44 witness45 statements46 and47 invented48 details49 about50 the51 shooter’s52 appearance.53 The54 filing,55 which56 was57 submitted58 on59 a60 Wednesday,61 said62 the63 brief64 included65 testimony66 from67 witnesses68 that69 do70 not71 exist72 and73 false74 descriptions75 of76 the77 clothing78 and79 features80 of81 the82 person83 accused84 of85 the86 shooting87.
During88 an89 August90 hearing,91 Justice92 C.93 Shannon94 Bacon95 confronted96 Aarons97 about98 his99 reliance100 on101 the102 AI103 system.104 She105 asked106 whether107 he108 follows109 news110 outlets,111 listens112 to113 the114 radio,115 or116 reads117 any118 coverage119 of120 current121 events,122 pointing123 out124 that125 problems126 with127 AI128 hallucinations129 appear130 regularly131 in132 mainstream133 coverage.134 Aarons135 acknowledged136 that137 he138 had139 used140 ChatGPT141 to142 help143 draft144 the145 brief,146 believing147 it148 would149 produce150 a151 reliable152 summary153 of154 the155 trial156 record.157 He158 said159 he160 did161 not162 realize163 the164 model165 could166 invent167 facts168 that169 have170 no171 basis172 in173 reality174.
The175 court’s176 order177 emphasized178 that179 attorneys180 have181 a182 duty183 to184 verify185 every186 factual187 claim188 and189 legal190 citation191 before192 submitting193 documents194 to195 the196 tribunal.197 It198 noted199 that200 the201 brief’s202 falsehoods203 undermined204 the205 integrity206 of207 the208 judicial209 process210 and211 could212 have213 misled214 the215 justices216 reviewing217 the218 appeal.219 The220 sanction221 serves222 as223 a224
Frequently asked questions
What penalty did New Mexico’s Supreme Court impose on lawyer Stephen Aarons for submitting AI-generated false testimony?
The court ordered Stephen Aarons to pay a $5,000 fine and found him in contempt after discovering fabricated witness statements and invented details in his appellate brief.
Why did the New Mexico Supreme Court sanction Stephen Aarons?
He submitted an appellate brief containing false witness statements and made-up descriptions of the shooter’s appearance that were produced by an AI tool, violating his duty to verify facts before filing.
What did Stephen Aarons say about his use of ChatGPT in drafting the brief?
Aarons acknowledged he used ChatGPT to help draft the brief, believing it would produce a reliable summary of the trial record, and said he did not realize the model could invent facts with no basis in reality.
What other legal cases involving AI-generated misinformation are mentioned in the article?
A federal judge reprimanded two law firms for false citations and quotations invented by AI, and attorneys for MyPillow’s founder were sanctioned for AI-generated misquotes and fabricated legal references in a court filing.
Source: The Verge
Related posts
Anthropic Files IPO to Beat OpenAI to Public Markets
Anthropic filed its S-1 registration with the SEC on October 26, planning to list on Nasdaq under ticker ANT and become the first major foundation model provider to go public, moving ahead of rival OpenAI. The filing reveals a $80.7 billion net loss on $16.3 billion revenue in the first half of 2026, largely from accounting charges, and roughly $369 billion in compute commitments over the next decade.
OpenAI Claims Three Scientific Breakthroughs With Academic Review
OpenAI announced three scientific breakthroughs-in mathematics, computer science, and physics-alongside a new "shepherd model" that enlists expert-led groups to verify results before public release. The claims include a proof of the local smoothing conjecture resolving Kakeya and Stein's restriction conjectures in all dimensions, a deterministic algorithm disproving the Exponential Time Hypothesis, and a rigorous existence proof for four-dimensional quantum field theories encompassing
Anthropic Files S-1 for IPO on Nasdaq
Anthropic filed a Form S-1 with the SEC on October 26, 2026, signaling plans for an initial public offering of Class A common shares on the Nasdaq Global Select Market under ticker ANT. The AI company, structured as a public benefit corporation, has not disclosed share count or price range. Goldman Sachs and Morgan Stanley are lead underwriters. The filing is preliminary and requires SEC approval before shares can trade.