Source: AI LAW BLOG

With its final approval on September 17, 2025, Italy has established its first comprehensive law on artificial intelligence, designed to integrate and specify the EU's AI Act. The national legislation is built upon the supremacy of European Union law, explicitly requiring that its interpretation conform to the AI Act. It also strictly forbids "gold plating"—the imposition of additional regulatory burdens on businesses beyond what the EU requires—to ensure a fully harmonized single market. To maintain consistency, the law adopts the complete legal glossary of the AI Act, ensuring that key terms share the same definition across the Union.

The law establishes a clear national governance structure, appointing the Agency for Digital Italy (AgID) as the notifying authority and the National Cybersecurity Agency (ACN) as the market surveillance authority. Beyond this foundational alignment, the Italian legislation introduces significant national provisions in areas not fully harmonized by the EU regulation. Key sector-specific rules have been created for healthcare, aiming to facilitate the use of data for scientific research, and for the judiciary, where the principle that AI can only assist, but never replace, a human judge is firmly established.

The most innovative aspects of the Italian law concern copyright and criminal liability. It amends existing copyright law to clarify that works are protected even if created "with the aid" of AI, provided there is a determining human creative contribution, and it strengthens the opt-out mechanism for Text and Data Mining. Crucially, the law introduces a new framework of criminal penalties, which the AI Act lacks. This includes a general aggravating circumstance for crimes committed using AI and creates a new, specific offense punishing the illicit dissemination of "deepfakes" with a prison sentence of one to five years. This ensures that while personal use is not subject to regulatory compliance, criminal abuse of the technology is severely punished.