This paper analyses the challenges arising from the integration of Artificial Intelligence (AI) in Italian public procurement, using judgement n. 4546 from the Lazio Regional Administrative Court on 3 March 2025, as subsequently affirmed by the Italian Supreme Administrative Court (Consiglio di Stato) with judgement n. 8092 of 20 October 2025, as a key case study. This ruling, the first in Italy to address the use of generative AI (ChatGPT-4) in a technical bid, serves as a lens to examine the adequacy of traditional legal frameworks. The analysis critiques the Court’s decision to dismiss the appeal by relying on the established principles of the evaluation committee’s broad technical discretion and the limited judicial review of assessments made using the Analytic Hierarchy Process (AHP) method. The author argues that this approach creates a “double shield”, making effective judicial review of the technical substance of AI solutions nearly impossible and leading to a “deference paradox”.