Source: UNC School of Law

UNC School of Law reports on a groundbreaking mock trial conducted on October 24, 2025, at the University of North Carolina at Chapel Hill, where three artificial intelligence systems—ChatGPT, Claude, and Grok—served as jurors. The experiment, designed by Professor Joseph Kennedy and part of the University's Converge-Con AI Festival, revisited a real juvenile robbery case where a human judge had previously convicted the defendant. In the simulation, set in a fictional 2036 under the "2035 AI Criminal Justice Act," the AI jurors unanimously acquitted the 17-year-old Black defendant, Henry Justus, citing insufficient evidence of shared criminal intent regarding the robbery committed by another student.

The AI systems demonstrated "strikingly human-like reasoning" during deliberations, with ChatGPT initially leaning towards conviction before being persuaded by the others to acquit based on the lack of clear encouragement or conduct. The experiment highlighted a sharp contrast between the AI's strict adherence to the burden of proof and the "imperfect" human judgment that led to the original conviction. Interim Dean Andy Hessick noted that while humans have biases and mental shortcuts, the AI verdict raises complex and unresolved questions about whether machines can or should ever replace human decision-makers in criminal proceedings.