Jump to content

Italian public-sector AI adoption guidelines (2026)

From AI Law Wiki

AI-generated text. This page was generated using artificial intelligence.

Law & policy information
Name
Guidelines for the Adoption of AI in Public Administration.
Jurisdiction
Italy.
Instrument type
Formal technical regulatory act under the Digital Administration Code, with mandatory and advisory provisions.
Bill / legal citation
AgID Determination No. 169 of November 9, 2026.[1]
Legislative or adoption status
Final version adopted and published online November 9. Gazette notice November 16, 2026.[1][2]
Effective dates
November 10, 2026, under Article 8 of AgID’s publication regulation.[3]
Last source verification
November 18, 2026 — formal status, publication and selected substantive provisions.

The Agency for Digital Italy (AgID) adopted and published the final Guidelines for the Adoption of AI in Public Administration on November 9, 2026, through Determination No. 169. AgID’s publication rule made them effective November 10. The November 16 Gazette carried a notice of the adoption. The guidelines apply to the entities covered by Article 2(2) of the Digital Administration Code and to AI systems and models within the definitions in the EU AI Act and Italy’s Law No. 132/2025.[1][3][2][4]

Formal status and reading rules

The guidelines follow the statutory CAD Article 71 procedure. AgID’s regulation on the digital ombudsman defines guidelines issued under that article as technical regulatory acts with generalized, binding effect.[5] Section 2.1 of the AI annex distinguishes mandatory requirements and prohibitions from recommendations and optional measures. Those distinctions govern how each provision is read.[6]

Responsibility and monitoring

Public bodies must assign responsibility for managing AI to the official responsible for digital transition and that official’s office. They must identify the risk level and their role under the AI Act, such as provider or deployer, and maintain risk management suited to the system and its use. Providers and deployers must monitor systems throughout their lifecycle, including their impact on fundamental rights. The guidelines call for documentation and personal-data checks even for systems at limited or minimal risk. For minimal-risk systems, they recommend a process to assess potential effects on fundamental rights, which can be simplified.[7]

Notices and review for affected people

Public bodies must explain why they use a system, the criteria it uses in decisions and its potential impact, such as on access to a service or benefit. Where legally permitted automated decisions significantly affect people, the guidelines detail safeguards under applicable law, including the GDPR when personal data is involved. Notices must explain available procedures and channels to contest a decision and request human review, including response times. The text calls for explanations of the logic and principal factors in a covered automated decision, subject to protections for other people, system security and trade secrets.[8]

The guidelines apply alongside the EU AI Act, Italy’s AI law and applicable data-protection rules. Italy’s Legislative Decree No. 179/2026 separately addresses supervision, enforcement and training. AgID reports that comments on its separate AI development and procurement drafts are under review following consultation.[9][10] See Italy AI Law and Government use and procurement.

References