Loading…
Loading…
Legge 132/2025 is the first national-law implementation of an EU-wide AI framework by any EU member state. Where Regulation (EU) 2024/1689 (the EU AI Act) sets the risk-based product-safety perimeter for AI systems placed on the EU single market, Legge 132/2025 layers on the institutional, sectoral, criminal-law and public-investment dimensions that the EU regulation deliberately left to member-state competence.
Architecture (6 chapters / 28 articles):
1. Principles (Chapter I) — human-centred, anti-discrimination, transparency, accountability principles framing AI research, experimentation, development and application within the Italian legal order. These are interpretive principles that bind sectoral secondary legislation. 2. National authorities (Chapter II) — designates AgID (Agenzia per l'Italia Digitale / Agency for Digital Italy) and ACN (Agenzia per la Cybersicurezza Nazionale / National Cybersecurity Agency) as Italy's national competent authorities under the EU AI Act, with policy coordination at the Presidency of the Council of Ministers (PCM) via a dedicated inter-ministerial committee. 3. Sectoral rules (Chapter III) — - Healthcare: patients have a right to know when AI is used in their care; non-discrimination obligations; the physician remains the ultimate decisional authority. - Labour & intellectual professions: AI use must be disclosed to workers and clients of regulated professions (avvocati, commercialisti, ingegneri, medici, etc.). - Public administration & judiciary: instrumental use of AI permitted but final decisions must rest with human officials/ magistrates; transparency and traceability obligations. - Minors: mandatory parental consent for AI services used by under-14s. 4. State investment (Chapter IV) — authorises up to €1 billion of investment via CDP Venture Capital (the state-backed VC arm of Cassa Depositi e Prestiti) into Italian AI, cybersecurity and telecoms companies, alongside aligned commitments for quantum and advanced computing. 5. Criminal & copyright (Chapter V) — new criminal offence for the malicious dissemination of AI-generated deepfake content capable of causing unjust harm, punishable by up to five years' imprisonment; modifications to copyright law to clarify the status of AI-generated and AI-assisted works. 6. Government delegations (Chapter VI) — broad enabling clauses empowering the Government to issue legislative decrees on AI liability, AI in the workplace, AI procurement standards, and alignment with downstream EU AI Act secondary acts.
EU AI Act perimeter, Italy's implementing architecture sets a reference template that other member states (France, Germany, Spain) will likely benchmark against. Squire Patton Boggs and Norton Rose both flag the Italian model as a leading early indicator for how EU national AI laws will diverge from each other within the Regulation 2024/1689 perimeter.
joins France's Bpifrance AI envelopes and Germany's AI strategy funding as third-largest EU national AI capital allocation by a member state. Materially increases pull on EU GPU capacity and on EuroHPC partner-utilisation slots.
G7 jurisdictions to attach explicit criminal liability (up to 5y) to harmful deepfake dissemination, ahead of EU-level harmonisation via the proposed AI Liability Directive. Sets a precedent that raises the policy floor for other member states.
services and public-administration AI deployments by foreign providers (US, UK) into the Italian market now face Italian-law obligations layered on top of the EU AI Act — disclosure regimes, human-in-the-loop requirements, parental-consent flows for under-14 user-facing products.
IPTM register entries (2023-08-10 Decreto Asset golden-power expansion, 2024-01-11 Piano Mattei Africa, 2026-XX-XX Legge 4/2026 golden-power financial-sector) and zero AI/digital regulatory instruments despite being the EU's third-largest economy and the first member state to enact a national AI law.
Chapter VI delegations give the Government up to 12 months from entry into force (i.e. by October 2026) to issue secondary legislation — those decrees will determine the operational bite of the statute (AI procurement standards, sectoral compliance templates, liability rules).
rules of Regulation 2024/1689 apply from 2 August 2025 and the high-risk system rules from 2 August 2026 — to what extent does AgID/ACN designation here pre-empt or coordinate with the EU AI Office's GPAI-model oversight?
envelope be drawn down, and what fraction will go to Italian vs cross-EU vs non-EU AI startups? Material for tracking the sovereign-AI investment flow into European compute and model development.