AI Policy in Italy: Key Provisions and What You Need to Know

Italy is gearing up to establish its own legislation on AI, which would complement the European AI Act. As reiterated during the June Italian-chaired G7 Summit, AI is a priority for Prime Minister Meloni's government, with the country "committed to developing governance mechanisms and ensuring that AI is both human-centred and human-controlled."

Despite the harmonising intent of EU-level legislation, the country is working on an "Italian Way" to AI that will have interesting implications for companies. The bill specifically targets the use of AI in healthcare, workplace, and copyright. Businesses in these sectors would have to comply with both EU regulation and Italian regulation going forward.

Understanding AI Policy in Italy: S. 1146 and its Implications

Italy has been swept up by the rapid changes in AI, with this technology described by Meloni as "the greatest revolution of our time." The country seeks to take advantage of AI, as emphasised by the high legislative activity in the area - 15 bills await analysis in the Italian Parliament. The government presented its version of AI policy with bill S. 1146, which was approved by the Council of Ministers.

Initially announced by Meloni at the event "Artificial Intelligence for Italy" in March 2024, the bill intends to represent the "Italian way" to AI, complementary to the European AI Act. AI represents "the main challenge we face from an anthropological, productive, economic and social point of view, [as there is the] danger of human intellect being superseded by it," said Meloni. She further announced a first €1 billion investment tranche, an impressive amount that has however been described by many as inadequate.

Healthcare

AI in healthcare plays a big role in this bill. According to the text, AI should allow for an overall improvement of the National Health System and of the living conditions of persons with disabilities, as well as provide support for disease prevention, diagnosis, and treatment. The anthropocentric view of technology in the healthcare sector would further be reiterated by

1. The need to periodically verify and update AI systems and their related data used in the healthcare sector, to ensure their reliability;

2. Patients' rights to be informed about the use of AI technologies, the diagnostic and therapeutic benefits derived from their use, and on the decision-making logic used; and

3. Deferring all decisions to medical professionals without prejudice.

The Italian bill appears to be treating all AI systems with the same level of risk. With the use of AI in the healthcare sector already regulated by EU-level legislation (AI Act, the European Medical Device Regulation - MDR, and the European Health Data Space - EHDS), these provisions are likely to make regulatory compliance harder for companies, translating into higher costs and time.

The picture could become even more complicated by the implementation of Law No. 86/2024, the so-called "differentiated autonomy," which intends to grant Italian regions additional special forms and conditions of autonomy on specific matters, currently of concurrent competence. This includes healthcare.

Workplace

Another area of legislative activity is the use of AI in the workplace. In line with the AI Act, Italy's bill states it should be used to improve working conditions, protect workers' integrity, and increase productivity. Expanding from the foundation of the AI Act, the bill would further regulate AI in work settings by introducing provisions on intellectual professions and for the Public Administration, as well as create an ad-hoc observatory within the Ministry of Work and Social Policies (MLPS). The latter would monitor the use of AI in the labour market, and propose AI-related policies and strategies.

The use of AI systems in the intellectual professions would be solely permitted to execute instrumental and support activities. Finally, employers or contractors would be obliged to inform workers of the use of AI systems in the workplace.

Companies operating in Italy should therefore ensure their compliance with EU- and national-level legislation, a task even more daunting as the bill does not distinguish AI systems on a risk-based basis. This will likely lead to excessive compliance burdens on employers. A similar risk could also stem from the delegation power the bill would confer to the government for the introduction of new criminal offences for the "failure to adopt or adapt safety measures for the production, circulation, and professional use of AI."

Investments

The bill would allegedly introduce investments totaling €1 billion in the fields of AI, cybersecurity, quantum computing, telecommunications, and their enabling technologies (Art. 21). However, the amount it refers to had already been established under the 5-year CdP Venture Capital Industrial Plan 2024-2028. What the bill would effectively introduce (Art. 19) is a maximum expenditure of €300.000 per year for 2025 and 2026 for the implementation of experimental projects on AI, related to the services provided by the Ministry of Foreign Affairs (MAECI).

While the measure has been largely portrayed by the executive as the clear proof of its focus toward AI, others are stating the invested funds are insufficient, and "unmask the intention of the government." The final article of the bill introduces a financial invariance clause, a formula to indicate that the implementation of the law should not translate into new or increased burdens on public finance, except for Art. 19. Instead, Art. 21 would autorise CdP Venture Capital to take participation in risk capital directly or indirectly in startups and innovative SMEs based in Italy and operating in the key sectors, or highly innovative enterprises "with high development potential" having their legal and operational headquarters in Italy.

The Italian association of leading ICT companies Anitec-Assinform argues that the allocated resources are not sufficient to ensure the competitiveness of Italy internationally, even more so compared to other European countries. Similarly, MP and Vice-President of the Upper House Anna Ascani described the bill as a "missed opportunity for Italy," a "no-cost measure" showcasing that "the government does not consider AI a priority [as,] if you are aware of the importance of a certain area, you will find the money."

Delegation Powers

The bill would empower the government to bring national law in line with the EU Regulation in matters such as AI literacy for citizens (in both school and university courses) and training by professional bodies for professionals and practitioners.

The delegation also covers reorganisation in criminal matters to adapt crimes and penalties to the illicit use of AI systems. It would also appoint AgID and ACN, both reporting to the Presidency of the Council, as the Italian competent authorities to supervise and enforce the AI Act's application and implementation. In late March, the Italian Data Protection Authority (GPDP) had already proposed itself for the task, as possessing the necessary competence and independence to implement the EU regulation.

Users' Protection, Watermarks & Copyright

Among the final articles of the bill are some of its potentially most impactful provisions, as they would regulate content sharing by audiovisual and radio services, including video-on-demand and streaming services.

To start, AI would be explicitly added to the country's Consolidated Text on Audiovisual Media Services (TUSMA) among the means and techniques capable of manipulating content in a way not recognisable by users. A further amendment to the TUSMA would concern the introduction of watermark provisions for textual, photographic, audiovisual, and radio content using AI systems.

Numerous opposition bills (S. 908, S. 917, S. 964, S. 965, and C. 1650) on the topic have been introduced in parliament. However, given the overarching nature of the government-sponsored AI bill, the latter is expected to be prioritised over any other proposal on the matter, with the possibility of these being joined at the committee stage for analysis.

Finally, providers of video-sharing platforms subject to Italian jurisdiction would have to take the necessary measures to protect the general public from content that has been fully or in part generated or altered using AI. They would further need to create functionality allowing users uploading content to declare if it has been AI-created or manipulated. Anitec-Assinform warned that this provision could isolate the country, as it would introduce a detailed provision for companies to comply with, while the AI Act yet has not.

On copyright protection, the bill would clarify the rules for AI-generated works. The Italian Copyright Law would be amended to include under its scope intellectual works "created with the support of AI tools, provided they are the result of the intellectual work of the author," and provisions on text and data mining (TDM), with the latter allowed to train AI models. Again, these practices are already regulated at the EU level (Directive on Copyright in the Digital Single Market, AI Act), with the possibility for copyright holders to prevent TDM practices. In this case, the national-level legislation would introduce a possibility otherwise discouraged.

Keeping Up With AI Policy in Italy and Around the World

In the ever-evolving landscape of technology regulation, Italy's AI bill has sparked diverse perspectives and discussions among experts and politicians. As it makes its way through the legislative process, the bill remains open to changes and refinements that could significantly alter its provisions. It's crucial to keep a close eye on its development.

In its current form, the bill aligns with the principles of the EU AI Act while it is also poised to introduce fresh compliance burdens for companies navigating the Italian market. The focus of the Italian government on AI, further highlighted by its inclusion in the G7 Summit agenda, suggests the willingness to lead the debate domestically and internationally. The country's unique approach to AI, the "Italian Way," is an interesting journey to follow. Yet, the proposed investments appear to be inadequate, likely limiting the expected impact of the proposal, and failing to place the country at the forefront of the AI race.

As Italy's national legislative framework evolves, proactive monitoring and strategic adaptation will be essential to navigate the shifting regulatory landscape effectively. FiscalNote's technology and expert-led insights can help you stay on top of further developments as the AI bill shapes the future of technology governance in Italy. The journey ahead promises to be dynamic and transformative, requiring a collaborative effort to ensure successful compliance and innovation integration.