Italy’s Council of Ministers has adopted a suite of legislative decrees that embed the European Union’s AI Act into national law, making the country the first to craft a coherent domestic framework that puts human oversight squarely at the centre of artificial intelligence. The package, approved in preliminary form, spans policing, employment, education, justice and public administration, but its most consequential provisions recalibrate how law enforcement may deploy algorithmic tools. Viewed from Brussels, the speed and ambition of Rome’s move signal a determination to shape the EU’s landmark regulation from the ground up, rather than passively transpose it.

For the police, the decrees break new ground: real-time remote biometric identification is authorised only in exceptional circumstances—terrorist threats, the hunt for fugitives or the search for victims of trafficking—and always requires prior judicial approval that fixes strict time and spatial limits. Even then, artificial intelligence remains a support tool, never an automated decision-maker. Interior Minister Matteo Piantedosi was at pains to stress that Italy is not building “a generalised Big Brother,” insisting that final choices will always belong to a human officer and that the use of vast biometric databases is expressly forbidden. Analysts in London note that these safeguards, while robust on paper, will be tested by the operational temptations of law enforcement; the requirement for “qualified human review and surveillance” raises immediate questions about training and institutional culture.

The government’s “anthropocentric” logic extends far beyond security. One decree explicitly prohibits algorithms from unilaterally deciding hiring or dismissals, cementing the principle that employment relations require human judgment. In universities and research institutions, three dedicated articles translate the same vision into concrete investment: AI must remain under human control and be oriented toward the common good, supported by funding for knowledge, training and responsible innovation. From Washington, where the regulatory debate remains fragmented, the Italian approach underscores a wider transatlantic divergence—Europe’s insistence on codifying human primacy before rolling out the technology.

The legislative package now faces parliamentary scrutiny, as well as review by the conference of regions and the national authorities that will enforce it. Forward-looking observers in Brussels see Italy’s detailed architecture as a potential template for other member states that are still wrestling with the AI Act’s implementation. Yet the real-world test will lie in how these norms interact with rapid technological change and the innate tension between civil liberties and the urge to harness predictive policing. For the moment, Rome has set down a marker: artificial intelligence must serve people, not the other way around.