AI Handled Cases Better Than Real Judges

At the University of Chicago Law School, 61 federal US judges were replaced by GPT-5 in a test on traffic accidents with different conditions. The result is impressive: AI applied the correct rule in 100% of cases, while humans made errors in about half of the cases.

GPT-5 completely ignored emotions and subjectivity, acting strictly according to the letter of the law.

Judges more often resorted to "gray areas" to achieve a fairer outcome, while AI did not budge an inch. The ideal formal judge — is this progress or a loss of human judgment?

By the way, in our bot Neuron, an even more advanced model 5.2 is available — in the Premium plans