Lawyer fined $5K over AI-hallucinated witnesses in a murder case

The Supreme Court of New Mexico has penalized a defense attorney for submitting an appellate brief that contained entirely fabricated witness accounts and false testimony generated by artificial intelligence.

The lawyer, Stephen Aarons, received a five-thousand-dollar fine and a contempt citation after failing to verify the accuracy of the document he submitted on behalf of a client convicted of murder. Aarons admitted to utilizing the popular chatbot ChatGPT to compile what he assumed would be a flawless and comprehensive overview of the trial proceedings, without realizing the tool's propensity to manufacture believable but completely fictional details. During a disciplinary hearing, the state supreme court justices sharply questioned how an experienced legal professional could remain so unaware of the widely reported risks, inaccuracies, and limitations associated with employing generative AI in high-stakes legal practice.

This disciplinary action highlights a growing challenge within the legal sector as more practitioners integrate automated software tools into their daily research and drafting workflows. Courts across the United States have increasingly encountered filings containing simulated citations, misquoted legal precedents, and fictitious evidence produced by hallucinating software programs. Similar penalties and reprimands have recently been handed down to other legal teams and high-profile representatives who failed to execute due diligence before filing automated briefs in federal and state courts. Aarons has publicly expressed deep regret over the oversight, characterizing the submission as an honest but careless mistake, while the broader judicial system continues to grapple with establishing strict guidelines and ethical guardrails to prevent the misuse of emerging technology in the courtroom.