UNC Law's AI justices match a Supreme Court ruling, not its feel

Three ChatGPT-based AI justices reached the same result as the US Supreme Court during an October 22 mock argument at the University of North Carolina School of Law. The outcome matched; the give-and-take did not, Reuters reported October 28.

A 2-1 vote that mirrored a 5-4 decision

At “May It Please the Machine,” third-year students Elijah Baetz and Reda Ansar argued Monsalvo Velázquez v. Bondi, a 2025 immigration case, UNC Law said. The issue: if a 60-day voluntary-departure deadline fell on Saturday, did it extend to Monday? That deadline marks when someone allowed to leave the US on their own must depart to avoid penalties.

Data librarian Rolando Rodriguez developed three AI justices using a single ChatGPT model. Each judge followed a different approach to interpreting the law. the textualist adhered very strictly to the literal meaning of the law. The pragmatist relied on practical and real-life results. On the other hand, the progressive considered a wider perspective, taking into account changing social values and modern legal principles. In order to ensure fairness and neutrality, all three worked only with anonymized legal records and remained offline while doing their work.

The extension was supported by the pragmatist and progressive in a ratio of 2-1, while the textualist dissented. In the actual Supreme Court decision, there was a 5-4 split with Justice Neil Gorsuch writing for the majority and Justices Clarence Thomas and Samuel Alito dissenting.

The machines reasoned but did not interrupt

The AI justices posed questions and presented possible situations, but they never interfered with the lawyers. Nicole Downing, the assistant dean for AI at UNC Law said the purpose of this was to test the capability of AI in reasoning like a judge does in written opinions, not to replace a judge. Clinical professor Annie Scardulla explained that oral argument is actually a “professional conversation.” That natural back-and-forth is missing in the AI justices.

Law professors rated AI tutors higher than peers

AI appears more promising as a tool for teaching law students than for making judicial decisions. In a blinded study, 16 contract professors from 14 US law schools reviewed 2,918 comparisons. They preferred AI-written answers 75.33% of the time. Only 3.53% of AI responses were flagged as harmful, compared with 12.06% of human-written answers. Still, helping students learn the law is one thing. Deciding cases is another.

Consumers draw a line at AI decisions

There’s a certain skepticism among the public too. Clio, in its 2026 survey, found out that only 26% of US consumers were willing to let AI make choices. If the matter was highly personal, this percentage dropped to 6%. Additionally, 82% of the participants wished to be made aware by their lawyers whether or not AI was being used.

What the trust gap means for legal AI vendors

Grand View Research values legal AI at $1.4 billion in 2024 and projects $3.9 billion by 2030, a 17.3% annual growth rate for 2025–2030. North America held 46% of 2024 revenue. Demand centers on research, document review and compliance, not automated judging.

Thomson Reuters found 77% of clients considered AI-enabled quality improvements very important or essential, but only 5% said most providers delivered them. Meanwhile, 34% of law firm professionals used unapproved tools.

Accuracy remains uneven: on a multi-state statutory-research benchmark, Westlaw AI scored 58%, Lexis+ AI 64%, and custom system STARA 83%. Those aren’t general legal-accuracy ratings.

As Cryptopolitan reported, Google launched Gemini Enterprise for Legal in preview, emphasizing traceable citations and client-data protections. UNC’s lesson is clear: AI can help lawyers reason, but humans must still judge.