Italian artificial intelligence research has gained a new place in the international debate thanks to ReTraceQA, a project presented by Sapienza University of Rome at the 64th Annual Conference of the Association for Computational Linguistics (ACL 2026), which was held in San Diego, California and is one of the most important global events for the natural language processing scientific community. The study, led by Roberto Navigli and Luca Moroni, Francesco Maria Molfese, Ciro Porcaro, and Simone Conia, addresses one of the most sensitive issues in the current era of generative AI: how much can we trust a model that provides the correct answer if the logical path that led to the solution is flawed? This is a critical matter for the reliability of artificial intelligence systems, thereby confirming the prominent position of Italian academic research in a field that is dominated by major international laboratories.