Italian AI law: A milestone, not a finish line

Italy has taken a bold step by approving a national AI law (Disposizioni e delega al Governo in materia di intelligenza artificiale) on the 17th of September 2025, placing itself among the first European countries to do so. The legislation with its 28 articles (official text waiting publication) serves as both a national framework and a bridge to the EU AI Act.

We unpack a few key features for you.

At its core, the bill embraces a "human-first" approach. It explicitly requires that AI systems respect human autonomy and decision-making power, ensuring transparency, explainability, and harm prevention. This also means humans, not algorithms, must remain accountable for final decisions, whether in healthcare, justice, or public administration. Judges, for example, must always retain the authority to interpret the law and assess evidence, even when AI is used in proceedings.

It places data governance at the heart of trustworthy AI, requiring that models be trained and operated on datasets that are accurate, reliable, secure, and fully auditable across the entire lifecycle.

To prevent misinformation and combat deepfakes, the law requires mandatory labeling of AI-generated content. Any audiovisual content created or significantly altered by AI must carry a clear "AI" marker. Digital platforms must also give users the option to declare AI use when uploading content.

With regard to safeguarding minors, the law requires parental consent for users under 14 years old for AI usage and clear messaging and accessible information to 14-17 year olds from AI product providers.

Cybersecurity The law requires safeguards to be embedded at every stage of the AI lifecycle. Oversight is shared between AgID (Agenzia per l’Italia digitale), which will handle innovation and conformity assessments, and the National Cybersecurity Agency (ACN- l’Agenzia per la cybersicurezza nazionale), which takes responsibility for supervision, inspections, and enforcement.

The bill clarifies copyright: works assisted by AI remain protected when they reflect human intellectual creativity, giving much-needed guidance to courts and creators navigating the complicated AI copyright landscape.

The law also provides sector specific guidance. For instance in the law’s research and public health provisions, the text anticipates the European Health Data Space (EHDS) regime and seeks to provide a legal base for using sensitive health data in AI research and innovation.

Looking Ahead

The Italian AI law is fresh and untested in practice. As with any sweeping framework, the real judgement will come when the law has to be applied and interpreted to specific use cases.

Key open questions include:

• What exactly qualifies as "significantly altered" AI content?
• What standard of audit evidence will satisfy regulators and courts?
• How will SMEs or regional health systems meet compliance without excessive burden?

This is just the start. Many details will be defined later through decrees and technical standards. The next months and years will reveal whether the law becomes a living governance system or an aspirational blueprint. Implementation will be the test.