# AI, Italy goes criminal > Source: > Published: 2026-08-05 08:15:58+00:00 **ROME —** Italy’s government has approved the final legislative decrees implementing the European Union’s AI Act, but Rome is doing more than simply transposing the bloc’s landmark regulation. **The package**, approved Tuesday by the Council of Ministers — the cabinet meeting chaired by Prime Minister Giorgia Meloni — introduces new national rules on the use of artificial intelligence in policing, education and, most notably, civil and criminal liability. **The result** is a distinctly Italian layer of regulation that industry observers say could reshape the legal risk facing companies developing or deploying high-risk AI systems in the country. **Why it matters: **Italy is among the first EU countries to accompany the AI Act with new criminal provisions targeting failures in AI safety and human oversight. - The move creates a liability framework that goes beyond the EU regulation, adding a national legal layer for companies operating in Italy. - Technology companies are closely watching the Italian approach as governments across Europe begin implementing the AI Act. - The debate comes as Italy is simultaneously trying to position itself as a destination for AI infrastructure and data-center investments. **The big picture: **The EU AI Act establishes a common regulatory framework across the European Union. But while the regulation is directly applicable, member states retain room to legislate in specific areas. Italy has chosen to use that space extensively. - Rather than limiting itself to implementing the European framework, Rome has built on its own national AI law (Law No. 132/2025), promoted by Undersecretary for Innovation Alessio Butti, creating an additional domestic regulatory architecture alongside the EU rules. - For international companies, that means compliance in Italy could involve not only meeting EU requirements but also navigating country-specific criminal and civil provisions. **Zoom in: Article 12. **The provision attracting the greatest attention is Article 12 of the implementing decree. - It introduces a new offence into Italy’s criminal code covering the failure to adopt technical safety measures or adequate human oversight for high-risk AI systems when such omissions create a danger to life or public safety. It also establishes criminal liability for unlawful alterations of high-risk AI systems and extends liability, under certain conditions, to cases of gross negligence. - According to a source familiar with the dossier, the real issue is not the AI Act itself but Italy’s decision to embed AI governance within its traditionally broad criminal liability framework. - “The real issue is Article 12,” a source said. ”Italy has a culture where criminal liability is everywhere.” **Between the lines: **The concerns emerging around the decree are less about whether AI should be regulated than about legal certainty. - Industry observers argue that companies may face criminal exposure before technical standards and practical guidance under the AI Act are fully consolidated. - That, they say, could make Italy an outlier within the European single market by exposing developers, executives and compliance officers to risks not found elsewhere in the EU. - Several large technology companies are already closely monitoring the Italian implementation, according to people familiar with the discussions, viewing it as an early test of how far individual member states may go beyond the common European framework. **The effect: **Some legal observers point to a broader characteristic of the Italian legal system: a tendency to extend personal liability up the corporate chain when safety failures are involved. - That perception has been reinforced by high-profile cases, including the final conviction of former Ferrovie dello Stato CEO Mauro Moretti over the 2009 Viareggio rail disaster, where senior executives were held personally accountable for systemic safety shortcomings. - While the case is unrelated to artificial intelligence, a source familiar with the AI debate said international technology executives are likely to factor this broader legal culture into their assessment of potential exposure in Italy. **Follow the money: **The timing highlights an apparent tension in Rome’s AI strategy. - In the same cabinet meeting that approved the AI liability framework, the government also advanced measures aimed at supporting AI infrastructure, including data-centers development. - The juxtaposition reflects Italy’s current balancing act: promoting itself as a hub for AI investment while simultaneously expanding the legal responsibilities associated with developing and deploying AI technologies. **Flashback: Remember the privacy! **This is not the first time Italy has taken a more interventionist approach to AI than most of its international peers. - In March 2023, shortly after ChatGPT went mainstream, Italy’s data protection authority became the first regulator to react at national level, temporarily blocking the service over concerns about the collection of personal data, age verification for minors and the lack of sufficiently clear information provided to users. - OpenAI made a series of changes requested by the regulator, and ChatGPT returned to the Italian market a few weeks later. - For some industry observers, that episode has become an early example of a broader regulatory approach in which Italian authorities are willing to move ahead of other jurisdictions, even if the practical outcome ultimately proves more limited than the initial intervention suggested. **What we’re watching: **Italy’s approach could become an early test case for how much room EU member states have to add national rules on top of the AI Act. - If other governments follow Rome’s lead, Italy may emerge as the standard-bearer of a more expansive interpretation of AI regulation — one that relies more heavily on criminal liability than the EU rulebook itself and that many technology companies are likely to scrutinize closely. - If, instead, Italy remains an outlier, the country could face questions over whether a more burdensome liability regime makes it a less attractive destination for AI investment and the deployment of advanced AI systems than other parts of the European market. **The bottom line: **Italy’s implementation of the AI Act is becoming more than a technical exercise in European compliance. - It is emerging as an early test of how far national governments are willing to extend liability beyond the EU rulebook. For companies deciding where to develop, deploy or invest in AI, the Italian question may increasingly be not only whether they can comply with the AI Act, but whether they are prepared for a legal environment that goes further than Brussels requires.