Italy

Article 20 designates the Agenzia per l'Italia digitale (AgID) and the Agenzia per la cybersicurezza nazionale (ACN) as Italy's national AI authorities, without displacing the Bank of Italy's, CONSOB's and IVASS's existing sector market-surveillance roles under EU AI Act Article 74(6) (comma 1).

AgID is responsible for promoting AI innovation and development and for defining the procedures and exercising the functions for notifying, assessing, accrediting and monitoring the bodies charged with verifying AI systems' conformity (comma 1, lettera a); ACN is responsible for supervising AI systems, including inspection and sanctioning activities, for cybersecurity compliance, and for promoting AI development on cybersecurity aspects (comma 1, lettera b); the two agencies jointly establish and manage AI regulatory-sandbox spaces, consulting the Ministry of Defence on dual-use AI and the Ministry of Justice on AI used in judicial activity (comma 1, lettera c).

Whoever causes unjust harm to a person by transferring, publishing, or otherwise disseminating, without that person's consent, an image, video, or voice recording falsified or altered through the use of an AI system and suited to deceive as to its genuineness, is punished with imprisonment from one to five years.

The offence is prosecutable only on the injured person's complaint, except that it is prosecuted ex officio if connected to another offence for which ex officio prosecution is required, or if committed against a person incapable by age or infirmity, or against a public authority because of its functions.

Article 4 requires that use of AI systems in employment and elsewhere ensure lawful, correct and transparent personal-data processing consistent with EU data-protection law, communicated in clear and simple language (commi 2-3); a minor under fourteen needs the consent of whoever holds parental responsibility to access AI technologies and for the resulting personal-data processing, while a minor aged fourteen to eighteen may consent alone if that information is easily accessible and understandable (comma 4).

Article 11 requires that AI used in the workplace be safe, reliable, transparent, respect human dignity, and not violate personal-data confidentiality, and requires an employer or client to inform the worker of AI's use in the cases and manner already set by Article 1-bis of Decreto Legislativo 152/1997 (the automated-decision-and-monitoring-system disclosure duty); AI used to organise or manage an employment relationship must observe the worker's inviolable rights without discrimination on sex, age, ethnic origin, religion, sexual orientation, political opinion, or personal, social or economic condition (commi 1-3).

Article 13 limits the use of AI systems within a regulated intellectual profession to instrumental and support activities, so that the professional's own intellectual work remains predominant in the service provided. To preserve the trust relationship between professional and client, it requires the professional to communicate to the client, clearly, simply and completely, the AI systems used.

The Garante's Provvedimento n. 146/2019 imposes specific security measures for genetic data (documented physical-access controls, encrypted or pseudonymized storage, controlled transmission), consent requirements for genetic testing and for processing genetic data to protect a third party's health, and a research-retention rule limiting secondary use of health data to cases where equivalent research cannot be done on data from consenting subjects.

The Garante's leading biometric enforcement action is Provvedimento n. 50 del 10 febbraio 2022 against Clearview AI: a EUR 20 million fine, an EU-wide processing ban on Italy-related facial-recognition data, and an order to delete data and designate an EU representative, for scraping and processing facial images with no valid legal basis.

No dedicated Garante decision or guidance on voiceprints specifically, or on employer fingerprint or facial-recognition timekeeping, is located; these are genuine gaps, not confirmed absences, since the Garante's own thematic search pages cannot be queried.

A controller must notify the Garante within 72 hours of becoming aware of a personal-data breach, and notify the affected individual without undue delay where the breach is likely to result in a high risk. No Italy-specific derogation from this timeline or threshold was identified in the Codice Privacy.

Italy gives the General Data Protection Regulation (GDPR) domestic effect through the Codice in materia di protezione dei dati personali (Personal Data Protection Code), Decreto Legislativo 196/2003 as amended by Decreto Legislativo 101/2018, in force from 19 September 2018. Beyond the GDPR baseline it adds its own Titolo III criminal offenses for unlawful processing (Artt.

167, 167-bis, 167-ter, 168, 170, 171; Art. 169 was abrogated outright by the 2018 decree), the Garante's own Article 166 sanctioning procedure, and sector-specific security and consent prescriptions for genetic, health, and biometric data.

A transfer of personal data outside the EEA requires an adequacy decision, Standard Contractual Clauses, Binding Corporate Rules, or a narrow Article 49 derogation, backed by the Article 83(5) top fine tier of up to EUR 20 million or 4% of global turnover. No Italy-specific derogation from this EU-wide framework was identified.

Individuals in Italy have the right not to be subject to a decision based solely on automated processing, including profiling, that produces legal or similarly significant effects. The Garante's ChatGPT/OpenAI matter is Italy's leading automated-processing enforcement episode: a provisional order suspending OpenAI's processing of Italian users' personal data on 30 March 2023, and a formal notice of violation on 29 January 2024.

A further EUR 15 million fine (December 2024) and the Court of Rome's annulment of that fine on a one-stop-shop competence ground (18 March 2026, once OpenAI's Irish establishment shifted lead-authority jurisdiction to Ireland's DPC) are not independently verified here, and rest on secondary sourcing rather than a confirmed read.

The Garante is Italy's supervisory authority, with General Data Protection Regulation (GDPR) Article 83 fines under the Codice Privacy Article 166 procedure.

GDPR Article 82 arms an individual with a direct private right of action, and Italy additionally has a general class-action mechanism, Codice di procedura civile Article 840-bis (azione di classe), in force since 19 May 2021, confirmed against the article's own text: it covers homogeneous individual rights against enterprises or public-service managers, brought by an individual class member or a registered nonprofit organization.

The article does not name data protection specifically, and whether it has actually been used for a GDPR claim is not established.

General Data Protection Regulation (GDPR) Article 9(1) lists biometric data processed for unique identification as a special category. The Codice Privacy adds no separate biometric-specific statute but does add its own Article 167(2) criminal offense specifically for unlawful Article 9-10 processing, and Provvedimento n. 146/2019's security and consent prescriptions for genetic and health data.

The Garante's Clearview AI decision confirms Italy applies no general publicly-available carve-out: processing publicly posted facial images to build a biometric identification database brought them within Article 9's special-category regime regardless of prior public availability.

Whoever, to procure an advantage for themselves or another or to cause harm to another, abusively obtains, holds, produces, reproduces, distributes, imports, communicates, delivers, otherwise makes available to others, or installs apparatus, instruments, parts of apparatus or instruments, codes, passwords or other means suited to accessing a computer or telematic system protected by security measures, or otherwise supplies indications or instructions suited to that purpose, is punished with imprisonment up to two years and a fine up to EUR 5,164 (comma 1).

The penalty rises to imprisonment from two to six years where one of the aggravating circumstances of art. 615-ter, comma 2, number 1) is present (comma 2), and to imprisonment from three to eight years where the conduct concerns the systems described in art. 615-ter, comma 3 (comma 3), both added when Legge 90/2024 amended commi 1-2 and inserted a new comma after comma 2, effective 17 July 2024.

Whoever abusively enters a computer or telematic system protected by security measures, or remains within it against the express or tacit will of whoever has the right to exclude them, is punished with imprisonment up to three years (comma 1); this base offence is prosecutable only on the victim's complaint.

Imprisonment rises to two to ten years where the offender is a public official or public-service employee abusing their powers or duties, or a person who also unlawfully practises as a private investigator, or an operator abusing their access to the system; where the offender uses threats or violence against people or property or is visibly armed; or where the offence causes destruction, damage, removal, inaccessibility, or interruption of the system, or destruction or damage of the data, information or programmes it holds (comma 2); these aggravated cases are prosecuted ex officio.

Where an offence under comma 1 or comma 2 reaches a system of military, public-order, public-security, health or civil-protection interest, imprisonment rises respectively to three to ten years or four to twelve years (commi 3-4). Legge 90/2024 raised the comma-2 introductory penalty range and amended comma 3, effective 17 July 2024; an earlier text of this article, reproduced in older secondary commentary, shows the comma-2 range as one to five years before that amendment.

Article 70-ter permits reproductions by research organisations and cultural-heritage-protection institutions, for scientific-research purposes, for text-and-data-mining (text and data mining (TDM)) from works or other material in networks or databases to which they have lawful access, and communication to the public of the research's results where expressed in new original works (comma 1); TDM is defined as any automated technique for analysing large quantities of text, sound, images, data or metadata in digital form to generate information, including patterns, trends and correlations (comma 2).

Article 70-quater permits, without prejudice to Article 70-ter, reproductions and extractions from works or other material in networks or databases to which there is lawful access, for TDM purposes generally, including commercial use, unless the rightholders of the copyright, related rights or database right have expressly reserved the use (comma 1); copies made under either article may be kept only as long as needed for the TDM purpose, and must be secured to a level at least equal to that required under Article 70-ter.

Legge 132/2025 (Italy's national AI law), art. 25, inserted Article 70-septies, confirming that reproductions and extractions for TDM through AI models and systems, including generative AI, from lawfully accessible works or database material, are permitted under the same Articles 70-ter and 70-quater conditions, subject to the Berne Convention, effective 10 October 2025.

The same law's art. 26(3) added a new criminal offence at Article 171, comma 1, lettera a-ter, punishing reproduction or extraction of text or data from works or material in networks or databases in violation of Articles 70-ter or 70-quater, including through an AI system; Article 171's chapeau states the penalty as a fine in Italian lire ("da L. 500 a L. 20.000") in the text, and no verified euro-converted figure is available, so that amount is not recorded here.

A database's maker ("costitutore"), meaning whoever makes a substantial investment, in money, time or work, to constitute, verify or present a database, holds the right, independent of any copyright protection the database may separately enjoy, to prohibit extraction or re-utilisation of the whole or a substantial part of its content (art. 102-bis, commi 1(a), 3).

"Extraction" is the permanent or temporary transfer of the whole or a substantial part of the database's content onto another medium; "re-utilisation" is any form of making that content available to the public (comma 1(b)-(c)).

The right arises when the database is completed and expires fifteen years after 1 January of the year following completion, or, for a database made available to the public before that period expires, fifteen years after 1 January of the year following first making-available; a substantial new investment in the database's content restarts an equally long protection term (commi 6-8).

Repeated and systematic extraction or re-utilisation of insubstantial parts of the database's content is not permitted where it presupposes operations contrary to the database's normal management or unjustifiably prejudices its maker (comma 9).

A legitimate user of a database made available to the public may not prejudice the rightholder, may not act contrary to the database's normal management or unjustifiably prejudice its maker, and may otherwise extract or re-use insubstantial parts, assessed qualitatively and quantitatively, of the database's content for any purpose without the maker's authorisation (art. 102-ter, commi 1-3). A contractual clause violating any of those rules is void (comma 4). Both articles have been in force since 16 June 1999.

Article 24 requires an essential or important entity, defined by Article 3 against the sector annexes and including the digital providers those annexes name (an online marketplace, an online search engine, a cloud computing service), to adopt technical, operational and organisational measures adequate and proportionate to the risks its network and information systems face, on a multi-risk approach covering at least risk-analysis and security policy, incident handling (including the procedures for the Article 25 and 26 notifications), business continuity and disaster recovery, and supply-chain security, transposing NIS2 Article 21.

Article 23 places approval and oversight of these measures on the entity's own management body, whose members can be held liable for an infringement under Article 38.

Article 25 requires an essential or important entity to notify CSIRT Italia, without unjustified delay, of any incident with a significant impact on the provision of its services, on a graduated clock: a pre-notification within 24 hours of becoming aware of the incident, a full notification within 72 hours, an intermediate report on CSIRT Italia's request, and a final report within one month of the notification (or, for an ongoing incident, monthly progress reports and a final report within one month of its resolution).

A trust-service provider notifies within 24 hours rather than 72. CSIRT Italia in turn responds within 24 hours of the pre-notification with an initial assessment and, on request, technical guidance, transposing NIS2 Article 23.

Establishes a legal duty for website operators and video sharing platform providers that distribute pornographic images or videos in Italy to verify that users are adults before granting access, delegating the technical and process details to AGCOM.

Adopts the binding technical and process rules that website managers and video-sharing platform providers must use to verify users are adults before granting access to pornographic content, using a dual anonymity model with independent certified third party verifiers so that neither the verifier nor the site learns both the user's identity and the service accessed.

Providers established in Italy had six months from publication, until 12 November 2025, to comply; AGCOM published a list of about 48 covered pornographic sites on 31 October 2025, and providers not established in Italy had until 1 February 2026.

Sets 14 as the minimum age at which a minor can independently consent to processing of personal data for an information society service, including registering for a social media account. Providers must obtain parental consent for users under 14. This functions as Italy's de facto minimum social media registration age, though it is a data protection consent rule rather than a mandated identity verification requirement.

Press publishers, whether acting singly, in association, or as a consortium, hold the exclusive reproduction and communication-to-the-public rights of LDA Articles 13 and 16 over the online use of their journalistic publications by information-society service providers, including media-monitoring and press-review businesses (comma 1).

A journalistic publication is a collection of mainly journalistic literary works, which may include photographs or video, forming a single item of a periodical publication under one title, published under an editor's or press agency's editorial responsibility and control; purely scientific or academic periodicals are excluded (comma 2). The right does not reach private or non-commercial use by individual users, hyperlinking, or the use of single words or very short extracts (comma 6).

A very short extract is defined as any portion of the publication that does not dispense the reader from consulting the article in full (comma 7).

An online service must pay the publisher equitable compensation ("equo compenso") for online use; AGCOM adopts the regulation setting the criteria for that compensation, weighing factors including the number of online consultations of the article, the publisher's years of activity and market relevance, the number of journalists it employs, both parties' technology and infrastructure investment costs, and both parties' economic benefits from the publication in visibility and advertising revenue (comma 8).

If the parties cannot agree an amount within thirty days of a request to negotiate, either may ask AGCOM to determine the equitable compensation; AGCOM decides within sixty days based on the same regulatory criteria (commi 9-10). If the parties still do not reach a contract after AGCOM's determination, either may bring the matter before the specialised business-law section of the ordinary courts (comma 11). Article 43-bis has been in force since 12 December 2021.

Current-affairs articles of an economic, political or religious character, published in magazines or newspapers, or broadcast, or made available to the public, and other material of the same character, may be freely reproduced or communicated to the public in other magazines or newspapers, including broadcast ones, if reproduction or use has not been expressly reserved, provided the source, date and author's name (if given) are indicated (comma 1).

Reproduction or communication to the public of protected works or material used on the occasion of current events is permitted for the exercise of the right to report news, within the limits of that informational purpose, again with source and author credited unless impossible (comma 2). The current text of Article 65 has been in force since 29 April 2003.

The summary, quotation or reproduction of passages or parts of a work, and their communication to the public, is free if done for criticism or discussion, within the limits justified by that purpose and provided it does not compete with the work's economic exploitation; if done for teaching or scientific research, the use must also be illustrative in purpose and non-commercial (comma 1).

A summary, quotation or reproduction must always be accompanied by the title of the work, the names of the author, publisher and, for a translation, the translator, where those appear on the reproduced work (comma 3). The current text of Article 70 has been in force since 9 February 2008.