The New Mexico Supreme Court has held a defence lawyer in contempt and fined him $5,000 after an appeal brief he submitted in a murder case contained fabricated police testimony and fictitious witnesses generated by ChatGPT.

The case involved attorney Stephen Aarons, who was representing Oscar Renee Sandoval in an appeal against his murder conviction. Sandoval had been convicted of killing the mother of his children. In preparing the appeal, Aarons used ChatGPT to summarise the trial proceedings, expecting the artificial intelligence tool to produce an accurate account of the evidence.

However, the resulting material included information that had never formed part of the trial record. According to the court, the AI-generated brief contained fabricated testimony attributed to police officers and statements from witnesses who did not exist. It also included inaccurate details concerning the alleged shooter's appearance.

The New Mexico Supreme Court criticised Aarons for submitting the material without independently checking it against the actual trial record. The justices also expressed disbelief at his claim that he was not fully aware of ChatGPT's capacity to generate false information, commonly referred to as AI "hallucinations."

STAR OF THE MONTH - Kwasi Amoako Adjei Esq.

The court held Aarons in contempt and imposed a $5,000 fine. The matter has also been referred to a disciplinary board for further investigation into his conduct as an attorney. The murder appeal has subsequently been reassigned to a public defender.

Aarons acknowledged that he had made a mistake and described the incident as a learning experience for legal professionals using artificial intelligence. His explanation, however, did not relieve him of his professional responsibility to ensure that material filed with the court was accurate.

AI tools can assist lawyers with tasks such as organising information, summarising lengthy materials and generating preliminary drafts. They cannot, however, replace the lawyer's obligation to verify facts, authorities, evidence and procedural history before relying on them in legal proceedings.

The case adds to a growing body of judicial responses to lawyers' uncritical reliance on generative AI in the US. Courts in several jurisdictions have already sanctioned legal professionals who submitted fictitious case authorities or other AI-generated material without verification.