Starting December 2, companies in Italy must comply with Article 5 of the European AI Act, as the deadline arrives for marking artificial intelligence systems already on the market.
This regulatory shift introduces financial risks, with authorities empowered to levy sanctions of up to 35 million euros for non-compliance, according to reports from ilroma.net.
Despite the approaching deadline, AI adoption and governance are not yet aligned. Data from the Politecnico di Milano’s Artificial Intelligence Observatory indicates that while 71% of large Italian enterprises had launched at least one AI project by 2025, only 9% established clear governance structures. Just 15% of these firms have active, structured projects to align with the European regulation.
“Before even considering the amount of potential fines, a company must know which AI systems it uses and in what capacity, whether as a provider or a user,” said Marco Panichi, an attorney specializing in AI law and corporate compliance, in an interview with ilroma.net.
Panichi said that the lack of internal expertise compounds the risk, pointing to a June Lenovo–Confindustria Giovani study where 82.4% of companies reported they lacked adequate internal AI skills.
Legislative Decree 179 Establishes Regulatory Oversight Roles
The regulatory framework is established by Legislative Decree No. 179 of October 7, 2026, which implements the European AI Act within Italy. Under this decree, the National Cybersecurity Agency (ACN) serves as the primary market surveillance authority and the central point of contact with the European Union, holding broad inspection powers.
Other institutions maintain specialized oversight. The Agency for Digital Italy (AgID) notifies the bodies responsible for assessing system conformity. Meanwhile, the Bank of Italy, CONSOB, and IVASS oversee high-risk systems within their respective financial and insurance sectors, and the Data Protection Authority retains its established role within the boundaries defined by the AI Act.
Penalty Tiers for Non-Compliance
Sanctions are categorized into four tiers, each tied to either a fixed monetary cap or a percentage of annual global turnover, whichever is greater. As reported by ilroma.net, the most severe penalties—reaching 35 million euros or 7% of turnover—apply to organizations utilizing prohibited practices.
Prohibited uses include software that recognizes employee emotions in the workplace, unless required for medical or safety purposes. Starting December 2, this tier also encompasses systems that generate intimate images of recognizable individuals without their consent.
Lesser penalties apply to other obligations:
- Up to 15 million euros or 3% of turnover: Applies to transparency violations, such as failing to inform customers they are interacting with a virtual assistant or neglecting to label AI-generated realistic content.
- Up to 7.5 million euros or 1% of turnover: Applies to providing inaccurate or incomplete information to authorities during investigations.
- Up to 1 million euros or 0.5% of turnover: Relates to fundamental rights impact assessments and the obligation to provide explanations for automated decisions, such as credit denials, starting December 2027.
For small and medium-sized enterprises (SMEs) and startups, the lower of the two penalty options applies. Panichi emphasized that the final figure depends on the severity and duration of the violation, whether the act was intentional or negligent, and the company’s cooperation with authorities.
Compliance Strategies
To meet these requirements, Panichi recommends that companies begin with a comprehensive census of all AI tools currently in use, including those purchased as services. Organizations must clearly define internal rules and ensure personnel receive training, as AI literacy is a mandatory requirement.
While some deadlines were shifted by Regulation (EU) 2026/1744, which came into force on July 27, 2026, others remain fixed. Transparency rules under Article 50 have applied since August 2, 2026. High-risk systems under Annex III, such as those used in education and essential services, face compliance deadlines starting December 2, 2027, followed by products regulated under Annex I on August 2, 2028. The mandatory AI literacy requirement, effective since February 2, 2025, remains unchanged.
Keep reading