L'IA può prevedere l'esito di un nuovo caso legale analizzando le sentenze dei giudici e i precedenti giuridici con una precisione del 90% ?

I modelli AI addestrati su migliaia di pareri giudiziari possono rilevare modelli di sentenze e interpretare argomentazioni legali complesse. Alcuni strumenti sono ora utilizzati nella strategia pre-processuale. La precisione diminuisce nelle giurisdizioni con dati scarsi o nuove teorie legali.

Background

AI models trained on thousands of court opinions can detect ruling patterns and interpret nuanced legal arguments; some tools are now used in pre-trial strategy. Accuracy drops in jurisdictions with sparse data or novel legal theories. Current AI systems assist in predicting legal outcomes by analyzing judge rulings, statutes, and precedents, but achieving 90% accuracy remains beyond current capabilities. Leading studies report accuracies in the 70–80% range for narrow, well-defined legal tasks, such as predicting outcomes in the European Court of Human Rights or U.S. Supreme Court cases, while broader or novel disputes introduce uncertainty that reduces reliability. These models rely on high-quality, annotated legal datasets and are most effective when applied to predictable jurisdictional patterns rather than unprecedented or complex fact patterns. The variability in judicial reasoning and evolving legal standards further limits consistent high-accuracy prediction. (Aletras, N., Vlachos, A., & Bengio, S, Enriched May 12, 2026)