AI Justices Mirror Supreme Court Decision in Mock Trial
The University of North Carolina experiment tested AI's legal reasoning to evaluate its utility as a courtroom tool.
Updated on Oct. 22, 2026 in Artificial Intelligence
Live Poll
Do you trust AI to perform accurate legal reasoning in professional settings?
The University of North Carolina School of Law recently tested the legal reasoning of artificial intelligence by holding a mock Supreme Court hearing. In the research-stage event titled May It Please the Machine, ChatGPT acted as the bench, reviewing case briefings and persona files without internet access.
Why it matters
This experiment evaluates whether large language models can replicate judicial reasoning to serve as potential legal support tools. By testing AI against historical Supreme Court outcomes, researchers aim to define the current limits of machine-based legal analysis.
The AI justices were evaluated with a performance score of 1.5/10 during the simulation. This assessment measured the system's ability to mirror judicial logic based on provided court briefings and specific justice persona files.
The players
University of North Carolina School of Law
An academic institution that focuses on legal research, pedagogy, and the application of new technologies in courtroom procedures.
ChatGPT
A generative pre-trained transformer model developed by OpenAI that uses deep learning to process text and simulate conversational or analytical responses.
The details
The simulation used ChatGPT in a disconnected state, restricting it from real-time internet searches to ensure its logic relied solely on pre-loaded data. Faculty provided the model with detailed court briefings and individual persona profiles to simulate the specific viewpoints of each Supreme Court justice. Students then presented oral arguments, and the AI system processed these to pose questions and eventually reach a verdict.
Timeline
October 22, 2026: The University of North Carolina School of Law hosted the mock oral arguments.
The Tech Race
This project represents a broader effort to quantify the efficacy of LLMs in highly structured, logic-heavy environments. It sits alongside ongoing industry attempts to determine whether AI can reliably serve as a drafting or analytical tool for legal practitioners.
Legal professionals in Chapel Hill and beyond should view these results as an indicator of current AI limitations rather than a ready-to-use tool for case preparation. The low performance score suggests that significant technical refinement is required before these models can be used for reliable legal reasoning.
The takeaway
The simulation demonstrates that while AI can reach the same conclusions as human justices, its current performance score indicates it lacks the depth required for complex legal deliberation. Watch for upcoming research from university legal departments on benchmarks for AI-assisted case analysis.
Further reading
For more on how machine learning intersects with modern litigation, visit Artificial Intelligence.
Live Poll
Do you trust AI to perform accurate legal reasoning in professional settings?
