EU AI Act in Italy: Law 132/2025, AgID, ACN, and What Companies Must Do
Italy's Law 132/2025 names ACN and AgID as AI Act authorities. Covers enforcement, penalties, GDPR overlap, and the Dec 2027 high-risk deadline.
Italy occupies a singular position in the EU's AI governance map. It is the first EU Member State to pass a national law dedicated to artificial intelligence — Law No. 132/2025 (Legge 23 settembre 2025, n. 132), in force since 10 October 2025. That law does not replace the EU AI Act; it cannot. Regulation (EU) 2024/1689 entered into force on 1 August 2024 and applies directly in every Member State by virtue of Article 288 TFEU. What Law 132/2025 does is build the national scaffolding: designating enforcement authorities, establishing principles, and delegating the Government to issue implementing decrees within twelve months.
For Italian companies, the practical result is a two-layer framework. The Regulation's obligations — what you must do, when, and at what fine ceiling — are set in Brussels. The question of who in Italy can investigate you, sanction you, and act as the country's interface with EU institutions is now answered by the national law.
A Regulation Plus a National Law: How They Fit Together
The EU AI Act is a Regulation, not a Directive. It does not require national transposition. Italian companies do not wait for an Italian implementation measure before their obligations bite — several already have. Article 5's prohibited practices have been enforceable since 2 February 2025. GPAI model obligations under Chapter V, the Article 99 penalty framework, and the governance structures at the AI Office in Brussels all applied from 2 August 2025.
Law 132/2025 operates within that pre-existing framework. It can designate authorities (which the Regulation requires Member States to do under Article 70), create domestic legal bases for inspections and sanctions, establish national principles for trustworthy AI development, and address intersections with other Italian law. It cannot alter the Regulation's obligation catalogue, change its deadlines, or soften its penalty ceilings. Those are fixed by Regulation (EU) 2024/1689 itself.
The implementing decrees that Law 132/2025 mandates — due within twelve months of 10 October 2025 — will add operational and technical detail to the national framework. Until they are published, the Regulation's own requirements govern, and the authorities designated by Article 20 of Law 132/2025 are already operational.
Italy's National AI Law: Law No. 132/2025
Law No. 132/2025, signed on 23 September 2025 and in force from 10 October 2025, is notable for two reasons beyond its content: it is the first national AI law in the EU, and it was enacted before most Member States had designated their Article 70 national competent authorities.
The law covers three broad areas. First, it establishes national principles for AI development and use — framing AI as a tool that must respect human dignity, non-discrimination, and fundamental rights, in alignment with but not duplicating the Regulation's own requirements. Second, and practically most important, it designates national competent authorities under Article 20 and establishes a high-level coordination body. Third, it delegates the Government to issue implementing decrees within twelve months, which will operationalise inspections, sanctions procedures, accreditation mechanisms for conformity assessment bodies, and technical guidance. Those decrees are pending as of mid-2026.
One aspect of Law 132/2025 that goes beyond the Regulation's administrative architecture: the law introduced national criminal provisions targeting harmful AI misuse — including, notably, the unlawful distribution of AI-generated deepfake content without the subject's consent. These are additions to Italy's criminal code, distinct from the Regulation's administrative fine regime. They apply to natural persons, not companies as such, and the specific provisions of Italy's criminal code that implement them are a matter of Italian domestic law rather than EU AI Act compliance — though companies with content or media functions should take note.
Who Enforces the EU AI Act in Italy?
Article 20 of Law 132/2025 designates two national competent authorities, with distinct but complementary mandates.
AgID (Agenzia per l'Italia Digitale — Italy's Digital Agency) holds the notifying-authority function. AgID is responsible for promoting AI development and innovation, and for the notification, assessment, accreditation, and ongoing monitoring of conformity assessment bodies (the bodies that conduct third-party conformity assessments for high-risk AI systems, primarily under Article 43). For Italian companies that need a notified body — mainly those building AI systems in the Annex III biometrics category, which requires the Annex VII notified-body route — AgID is the authority through which those bodies are accredited and notified to Brussels.
ACN (Agenzia per la Cybersicurezza Nazionale — National Cybersecurity Agency) is the market surveillance authority, the inspection authority, and the authority with sanctioning powers. ACN supervises AI systems on the Italian market, conducts inspections, can require access to technical documentation and conformity assessment records, issue corrective orders, and impose the administrative fines that flow from the Regulation's Article 99. ACN also serves as Italy's single point of contact (SPoC) with EU institutions — including the AI Office in Brussels — under Article 70 of the Regulation.
The assignment of market surveillance to ACN is deliberate. Italy's cybersecurity regulator already holds significant inspection and enforcement powers over digital infrastructure; extending that mandate to AI system supervision creates institutional continuity for the technical assessments involved, particularly for systems that intersect with critical infrastructure or cybersecurity (Annex III point 2).
Law 132/2025 also requires active collaboration between the two designated authorities and other Italian regulators where competences intersect. The Garante (Garante per la protezione dei dati personali — Italy's data protection authority) is not the primary AI Act authority, but the law mandates coordination with the Garante wherever AI system supervision touches personal data protection — which, in practice, covers nearly all Annex III high-risk systems. Similarly, AGCOM (Autorità per le Garanzie nelle Comunicazioni — Italy's communications regulator) must be brought in where AI systems involve communications, media, or content services.
A Coordination Committee at the Presidency of the Council of Ministers provides high-level oversight of AI policy and ensures coherence across the national framework. This body sits above the operational authorities; it does not itself conduct enforcement.
For GPAI model providers operating in Italy: the EU AI Office in Brussels supervises GPAI model obligations directly under Articles 53 and 55. ACN acts as SPoC, but the primary regulatory relationship is with Brussels, not Rome.
How Italy's Framework Interacts with the GDPR
The GDPR and the EU AI Act run in parallel for any AI system that processes personal data. For Annex III high-risk systems — recruitment screening, credit scoring, public-benefit eligibility assessment, biometric categorisation — that means both frameworks apply simultaneously.
The most practical overlap is between GDPR Article 35 and EU AI Act Article 27. GDPR Article 35 requires a Data Protection Impact Assessment (DPIA) before processing that is likely to result in a high risk to natural persons. EU AI Act Article 27 requires certain deployers to complete a Fundamental Rights Impact Assessment (FRIA) before putting a high-risk AI system into service. Article 27(4) explicitly recognises that the FRIA may build on an existing DPIA, sharing the same factual foundation. For an Italian public body — a municipality running an AI system to determine social-benefit eligibility, for example — conducting both a GDPR Article 35 DPIA and an Article 27 FRIA before deployment is mandatory, and the two assessments should be coordinated from the outset rather than treated as separate exercises.
A second interaction involves automated decision-making. GDPR Article 22 restricts solely automated decisions that produce legal or similarly significant effects for individuals, and requires meaningful information about the logic involved. EU AI Act Article 12 requires high-risk AI systems to maintain logs of their operation. The deployer running an AI creditworthiness assessment tool faces simultaneous obligations: log retention under Article 12 (at least six months under Article 26), documentation of automated-decision logic under GDPR Article 22, and human oversight under Article 14 of the AI Act. These obligations are distinct, but the documentation built for one informs the other.
The Garante's coordination role in Italy's framework means the data protection authority can flag compliance gaps to ACN and vice versa. Italian companies should not treat GDPR compliance and AI Act compliance as siloed programmes — an inadequate DPIA on a high-risk AI system is simultaneously a GDPR exposure and a signal that the FRIA under Article 27 may also be deficient.
The EU AI Act Timeline as It Applies in Italy
| Date | What applies |
|---|---|
| 1 August 2024 | Regulation (EU) 2024/1689 enters into force |
| 2 February 2025 | Article 5 prohibited practices and Article 4 AI literacy — enforceable now |
| 2 August 2025 | GPAI obligations (Chapter V, Articles 51–56), governance, AI Office, Article 99 penalties |
| 10 October 2025 | Law No. 132/2025 in force — AgID and ACN designated as national competent authorities |
| 2 August 2026 | General application including Article 50 limited-risk transparency (chatbots, deepfakes, synthetic-content marking) |
| 2 December 2027 | Stand-alone high-risk AI systems (Annex III list) — deferred under the Digital Omnibus |
| 2 August 2028 | High-risk AI as safety components of Annex I regulated products — deferred under the Digital Omnibus |
Two timeline points deserve particular attention for Italian businesses.
Article 5 is already enforceable. Any AI system that performs biometric categorisation by sensitive characteristics outside the Regulation's narrow exceptions, enables real-time remote biometric identification in public spaces beyond the law-enforcement carve-outs, scores individuals for social trustworthiness, or manipulates persons through subliminal techniques has been operating illegally since 2 February 2025. ACN already has authority to act.
The high-risk deadline is not August 2026. Under the Digital Omnibus — the Commission's amendment package for which Parliament and Council reached political agreement on 7 May 2026 — stand-alone Annex III high-risk systems now have until 2 December 2027, and high-risk AI embedded in Annex I regulated products has until 2 August 2028. The original August 2026 date has been deferred. That deferral does not make the documentation easier to build: an Article 9 risk management system, a full Annex IV technical documentation pack, Article 14 human oversight controls, and conformity assessment under Article 43 represent months of focused work for any provider.
Penalties: What Companies in Italy Face
The penalty framework is Article 99 of Regulation (EU) 2024/1689, enforced in Italy by ACN. There are three tiers:
- €35,000,000 or 7% of total worldwide annual turnover (whichever is higher) — for violations of the Article 5 prohibitions. This tier has applied since 2 August 2025.
- €15,000,000 or 3% — for non-compliance with most other obligations, including high-risk AI requirements under Articles 9–15, provider obligations under Article 16, deployer obligations under Article 26, and Article 50 transparency duties.
- €7,500,000 or 1% — for supplying incorrect, incomplete, or misleading information to notified bodies or competent authorities.
For smaller companies, Article 99(6) provides a proportionality protection: for SMEs and start-ups, the fine is capped at the lower of the fixed amount or the percentage. A company with €8 million in worldwide turnover cannot face a €15 million fine for a high-risk obligation breach — 3% of €8 million is €240,000, and that is the applicable ceiling.
Fines are maximums, not automatic outcomes. ACN will apply proportionality factors including the duration and nature of the infringement, the degree of responsibility, and cooperation with the investigation. But the ceilings are real, and the 7% tier for Article 5 violations is the highest administrative fine in EU regulatory law.
In addition to the Regulation's administrative fines, Law 132/2025 introduced national criminal provisions targeting specific harmful AI misuses — most notably the unlawful distribution of AI-generated deepfake content. These criminal provisions apply to individuals and sit alongside (not within) the Regulation's administrative penalty structure. Companies with content production, media, or communications functions should review Law 132/2025's criminal provisions as a separate compliance check, in addition to the administrative obligations imposed by the Regulation.
GPAI-specific fines are a separate instrument: up to €15 million or 3%, imposed by the Commission directly on GPAI model providers under Article 101.
Italy-Specific Compliance Considerations
Sandboxes and AgID's Role
Article 57 of the Regulation requires each Member State to establish at least one regulatory sandbox by 2 August 2026, with priority and free access for SMEs and start-ups under Article 58. AgID's mandate — promoting AI innovation alongside its notifying-authority function — positions it as the natural host for Italy's regulatory sandbox. The implementing decrees due under Law 132/2025 are expected to clarify the sandbox framework. Italian companies developing novel AI applications, particularly in sectors where real-world testing raises regulatory questions, should monitor AgID's guidance closely as the sandbox infrastructure takes shape.
Public-Sector FRIA Obligations
Italian public bodies — ministries, regional administrations, municipalities, public healthcare bodies — are among the most significant deployers of AI in Annex III categories: public-benefit eligibility, law enforcement support, border management, administrative justice. Article 27 of the Regulation makes the FRIA mandatory for public-body deployers before putting a high-risk AI system into service. The FRIA must be completed before deployment, documented, and made available for inspection by ACN. Italian public administrations should treat FRIA documentation as part of their procurement and deployment standard — and coordinate with the Garante on the overlapping GDPR Article 35 DPIA.
Manufacturing and the Annex I Product Route
Italy's industrial base — manufacturing, machinery, automation, industrial robotics — creates specific exposure to the Regulation's Annex I product safety-component route. Where AI is a safety component of machinery covered by the revised Machinery Regulation (EU) 2023/1230, it is high-risk under Article 6(1) rather than Annex III. The applicable deadline is 2 August 2028, but the conformity assessment pathway under Article 43 for Annex I products generally involves third-party notified bodies rather than self-assessment. Italian manufacturers embedding AI into safety-critical systems should: identify whether the Annex I or Annex III route applies; engage with notified bodies accredited by AgID early; and not assume that the 2 August 2028 date gives unlimited time — notified-body capacity is finite and demand is rising.
Deployer vs. Provider: The Article 25 Line
Most Italian companies deploying third-party AI tools sit in the deployer role under Article 26. Deployer obligations are real — human oversight, monitoring, log retention, Article 27 FRIA for qualifying deployers — but lighter than the provider stack. The risk lies in Article 25: a deployer who substantially modifies a high-risk AI system, places it on the market under their own name, or changes its intended purpose crosses into provider obligations under Article 16, which include the full Annex IV technical documentation, conformity assessment under Article 43, Declaration of Conformity under Article 47 / Annex V, and registration under Article 49. Italian software companies and system integrators who customise third-party AI products should conduct an Article 25 analysis before market launch.
How Confir Helps Companies in Italy
Italian compliance teams building their EU AI Act programmes face a documentation task that is simultaneously technical and legal: Article 9 risk management records, the nine-area Annex IV technical documentation pack, Article 27 FRIAs for qualifying deployers, conformity assessment preparation under Article 43, and an AI inventory that ties it all together.
Confir handles that documentation workload without consulting engagements or multi-month implementations. Its classification engine is rule-based and deterministic — it encodes Articles 5 and 6 with Annex III logic in explicit rules, so the same intake produces the same finding every time, with a human-readable explanation of which rule fired. That reproducibility matters for a compliance product: ACN's inspectors will expect to see the same answer in your documentation as in your assessment tool, and Confir's audit log makes that defensible.
Specifically, Confir generates the full Annex IV technical documentation pack (Article 11), the Article 47 / Annex V EU Declaration of Conformity, and runs the Article 27 FRIA. The structured assessment spans four areas mapped to specific Articles: risk classification and compliance (AIRC: Articles 5, 6, 43, 50), data and technical robustness (AITR: Articles 10, 11, 15), transparency and human oversight (AITO: Articles 13, 14, 27, 50), and governance and post-market monitoring (AIGM: Articles 9, 72, 73). EU-hosted, self-serve, from €600 per year.
What Companies in Italy Should Do Now
Immediately — Article 5 obligations are already enforceable. Audit any AI system that touches biometric categorisation by sensitive characteristics, real-time remote biometric identification in public spaces, social scoring, or manipulation techniques. If a system fits an Article 5 category without a statutory exception, it must stop or be restructured. ACN has authority to act now.
Before 2 August 2026 — Article 50 transparency. Any AI system that interacts with natural persons — chatbots, virtual assistants, synthetic-content generators, emotion-recognition tools — must comply with Article 50's disclosure obligations: users must be informed they are interacting with AI, synthetically generated content must be labelled, and emotion-recognition or biometric-categorisation uses must be disclosed. This is not the high-risk deadline; it is a separate obligation with its own enforcement ceiling.
2026 through 2027 — high-risk preparation for Annex III systems. The Digital Omnibus deferral gives stand-alone Annex III systems until 2 December 2027. Use 2026 to build your AI inventory, classify each system under Articles 5 and 6 (applying the Article 6(3) filter before assuming high-risk), assign provider and deployer roles accurately, and begin the Annex IV documentation process. For public-body deployers, initiate FRIA planning in parallel with any existing GDPR Article 35 DPIA programme.
Ongoing — monitor implementing decrees and ACN/AgID guidance. Law 132/2025's implementing decrees are due by October 2026. They will add operational detail on inspections, accreditation of conformity assessment bodies, and sandbox access. ACN and AgID are expected to publish guidance as those decrees take shape. Companies with Annex I product routes should engage with AgID-accredited notified bodies now, before capacity constraints tighten.
Frequently Asked Questions
Who enforces the EU AI Act in Italy?
ACN (Agenzia per la Cybersicurezza Nazionale) is Italy's market surveillance authority under Article 20 of Law 132/2025. ACN has inspection and sanctioning powers and serves as Italy's single point of contact with EU institutions. AgID (Agenzia per l'Italia Digitale) holds the notifying-authority function, responsible for accrediting and monitoring conformity assessment bodies. The EU AI Office in Brussels supervises GPAI model providers directly.
What is Law No. 132/2025?
Law No. 132/2025 (Legge 23 settembre 2025, n. 132) is Italy's national AI law, in force from 10 October 2025. It is the first comprehensive national AI law enacted by an EU Member State. It designates AgID and ACN as national competent authorities, establishes national AI principles, creates a coordination structure at the Presidency of the Council of Ministers, and delegates the Government to issue implementing decrees within twelve months. It complements Regulation (EU) 2024/1689 — it does not alter the Regulation's obligations or deadlines.
What do AgID and ACN each do under the EU AI Act?
AgID is the notifying authority: it assesses, accredits, and monitors conformity assessment bodies (the organisations that conduct third-party conformity assessments for high-risk AI, primarily for biometric systems under Annex VII). ACN is the market surveillance authority: it supervises AI systems on the Italian market, conducts inspections, requires access to technical documentation, issues corrective orders, and imposes fines under Article 99. ACN is also Italy's single point of contact with Brussels.
What are the EU AI Act fines in Italy?
The fine tiers under Article 99 apply uniformly across the EU. The highest tier — €35 million or 7% of worldwide annual turnover, whichever is higher — applies to Article 5 prohibition breaches. A middle tier of €15 million or 3% applies to most other obligations, including high-risk requirements and deployer and provider duties. A lower tier of €7.5 million or 1% applies to supplying incorrect or misleading information to authorities. Article 99(6) caps fines for SMEs and start-ups at the lower of the fixed amount or the percentage.
Does the Garante regulate AI in Italy?
The Garante (Italy's data protection authority) is not the primary EU AI Act authority. That role belongs to AgID and ACN. However, Law 132/2025 mandates active coordination between those authorities and the Garante wherever AI system supervision intersects with personal data protection — which covers nearly all Annex III high-risk systems. For companies, this means both the Garante (for GDPR compliance on high-risk AI) and ACN (for EU AI Act compliance) are relevant regulators, and gaps in one framework are likely to attract scrutiny from the other.
When do high-risk AI obligations apply in Italy?
Under the Digital Omnibus amendment (political agreement reached 7 May 2026), stand-alone high-risk AI systems listed in Annex III have until 2 December 2027. High-risk AI embedded in Annex I regulated products — machinery, medical devices, vehicles — has until 2 August 2028. Article 5 prohibitions apply now. Article 50 limited-risk transparency obligations apply from 2 August 2026. These dates are set by the Regulation and apply equally across all EU Member States, including Italy.
Does Italy's national law change what I need to do for EU AI Act compliance?
No. Your obligations — what to document, which assessments to run, when by, and at what fine ceiling — are set by Regulation (EU) 2024/1689. Law 132/2025 determines who enforces those obligations in Italy (ACN and AgID) and adds some national elements, including the coordination structure and the criminal provisions on deepfake content. The implementing decrees due by October 2026 will add procedural detail, but will not alter the Regulation's substantive requirements.
Related guides
- EU AI Act explained
- EU AI Act summary and key obligations
- EU AI Act in Germany: authorities and obligations
- EU AI Act penalties: the three tiers explained
- EU AI Act timeline and key deadlines
- Extraterritorial reach of the EU AI Act
- Annex III high-risk AI use cases
- AI risk classification under Articles 5 and 6
- Conformity assessment for high-risk AI
- Manufacturing AI compliance and the Annex I route
- Deployer obligations under Article 26
- Public-sector AI compliance and the FRIA
Manage your EU AI Act compliance in one place
Confir automates risk classification, technical documentation, and audit trails for any company. No consultants. No 6-month projects. 7-day free trial.
Start free trial →