The New Mexico Supreme Court held a lawyer who used ChatGPT to submit a brief with fabricated testimony in direct contempt of court.
The lawyer, Stephen Aarons, admitted to not verifying the factual claims and legal authority in his AI-generated brief before filing it with the court.
Aarons, a 40-year veteran criminal defense attorney, was fined $5,000 and barred from appearing before the court pending disciplinary proceedings.
Aarons submitted a brief containing false testimony from fabricated witnesses, including fake police statements and misrepresentations of real cases. The court found that he failed to verify the accuracy of the information, which led to the brief being struck from the record.
The court also ordered the public defender to appoint a new lawyer for the defendant and reset the case for the 2026-27 term.
Aarons admitted to feeding a computer-generated trial transcript into ChatGPT, which outputted the false quotes. He claimed he believed the AI would provide an accurate summary of the proceedings.
However, the court criticized him for not understanding that AI tools can generate false information. Justice C. Shannon Bacon noted that Aarons' claim of ignorance was inconsistent with the code of conduct.
Aarons argued that he did not know AI could hallucinate facts, but the justices pointed out that the issue had been widely reported for over three years. Bacon emphasized that the lawyer’s responsibility to verify information remains regardless of the source. The court stressed that the client, not the lawyer, is the one suffering from the errors in the brief.
The case highlights the risks of relying on AI without proper verification. Aarons suggested that the court issue a standing order requiring briefs to include a certificate of compliance regarding AI use.
However, the justices dismissed this as a distraction from the main issue: the lawyer’s failure to verify the brief’s accuracy. The disciplinary board will now determine further penalties.
Source: arstechnica
