Italy Italy AI strategy national framework

Italy's AI Act Decree Gives a Cybersecurity Agency the Market-Surveillance Role, and Proportionality Now Depends on How ACN Uses It

Legislative Decree 179/2026 makes ACN Italy's lead AI market regulator and AgID its notifying authority. Its value will depend on enforcement restraint and sandbox access.

Italy's AI Act Decree 179/2026 People of Internet Research · Italy €35M / 7% Max fine, prohibited practices Ceiling for Article 5 breaches und… €15M / 3% Max fine, operator duties Tier covering transparency and mos… 23 Oct 2026 Decree in force since Published in the Gazzetta Ufficial… peopleofinternet.com
Italy's AI Act Decree 179/2026 People of Internet Research · Italy €35M / 7% Max fine, prohibited pract… €15M / 3% Max fine, operator duties 23 Oct 2026 Decree in force since peopleofinternet.com

Key Takeaways

Italy has now finished the institutional half of its national AI framework. Legislative Decree No. 179/2026 was published in the Gazzetta Ufficiale on 8 October 2026 and entered into force on 23 October 2026. According to the IAPP's analysis, it is the second step after Law No. 132 of 23 September 2025. Law 132 set the principles. The decree decides who enforces the EU AI Act in Italy and with what powers.

What the decree does

The IAPP reports that the Agency for Digital Italy (AgID) is the notifying authority. It designs the procedures to assess, designate and notify conformity-assessment bodies. It then monitors them through periodic and extraordinary audits, and it can start proceedings to restrict, suspend or withdraw a body's designation.

The National Cybersecurity Agency (ACN) is the main market surveillance authority and Italy's single point of contact with EU institutions. Its powers include investigation and enforcement, and it can inspect real-world testing without prior notice. ACN also keeps the national register of high-risk systems used in critical infrastructure, which the AI Act lists in point 2 of Annex III.

A&O Shearman's summary adds three points. First, the Bank of Italy, CONSOB, IVASS and the data protection authority keep their supervisory powers in their own areas. Second, a Coordination Committee supports cooperation and data exchange among authorities. Third, the penalty ceilings follow the AI Act's tiers: up to €35 million or 7% of worldwide turnover for prohibited practices, €15 million or 3% for most operator obligations, and €7.5 million or 1% for supplying misleading information to authorities.

The strongest case for this design

The case for this design is real. The AI Act, Regulation (EU) 2024/1689, only works if each member state has an authority that can act. A regulation without a named enforcer produces forum-shopping and uneven protection. Putting the single point of contact in one agency gives Brussels and foreign providers one address. Splitting notification (AgID) from surveillance (ACN) also keeps the body that accredits auditors apart from the body that polices the market. Keeping financial supervisors in place avoids a new generalist regulator second-guessing credit and insurance models that Italian supervisors already understand. The decree is a plain, competent implementation of an EU obligation.

Where proportionality is at risk

The design still carries three risks that a pro-innovation reading should name.

First, the regulator's institutional culture. ACN is a security agency. Security regulators are used to treating risk as something to eliminate, so they can lean toward preventive controls. AI deployment in Italian industry, such as logistics software, public-sector tools and start-ups building on foundation models, is mostly ordinary commercial software. If ACN treats every high-risk classification like a critical-infrastructure incident, compliance costs will fall hardest on small firms. Large providers already employ conformity teams. The decree text matters less than how ACN writes its guidance and picks its first cases.

Second, the employment rule. According to A&O Shearman, the decree requires final decisions on hiring, modification or termination of employment, including discipline and performance reviews, to rest with a natural person who has genuine decision-making authority. Non-compliant dismissals are void. Human accountability for firing decisions is defensible. But "genuine decision-making authority" is a standard that courts will have to define case by case, and a void dismissal is a heavy consequence for an ambiguous test. Employers will respond by stripping useful decision support out of HR tools rather than risking a voided decision. That would not make workers any safer.

Third, penalties without safe harbours. Fines are set by the EU tiers, so Italy had little room to move on the ceilings. The fines that start at the top of the range are for prohibited practices, and they are relatively easy to avoid. The tier that will actually bite small firms is the one for transparency and operator obligations. Enforcement restraint, such as warnings before fines and published enforcement criteria, is where national discretion remains. The decree does not guarantee it.

What to watch

Three things will show whether Italy's framework supports innovation or only polices it.

The decree also leaves a general-purpose question open. A single national regulator can make a sound call on a bank's credit model or a municipal chatbot. It cannot, alone, settle how Europe regulates frontier models. Italian authorities should therefore align closely with the EU AI Office and avoid adding national obligations on top of the EU text. Law 132/2025, which Normattiva records as published in Gazzetta Ufficiale n. 223 of 25 September 2025 and in force from 10 October 2025, already gave Italy national additions. Decree 179 should not become the second layer of gold-plating.

Bottom line

Italy has built a workable enforcement structure quickly, and it deserves credit for doing so. The structure is only a starting point. A cybersecurity agency now holds the lead market-surveillance role. Its restraint in the first year, its guidance and the speed of its designations will decide whether the decree helps Italian AI firms or taxes them.

Sources & Citations

  1. IAPP: Italy's AI Act framework takes shape
  2. A&O Shearman: AI Act, Italy's implementing legislation
  3. Normattiva: Law 132/2025 on artificial intelligence
  4. EUR-Lex: Regulation (EU) 2024/1689 (AI Act)