ItalyDecreto Legislativo 160/2026, Police Use of AI Systems: Human Review, Oversight, Collaborations, Sandbox Basis and Training (Artt. 3-6, 21)
D.Lgs. 9 settembre 2026, n. 160, artt. 3-6 e 21
In force since .
An AI risk obligations rule binding government bodies.
- Obligation class
- Governance
As of .
What it requires
- These articles bind the Italian police forces and set conditions on AI systems and models developed, tested or used for police activity (artt. 3-6).
- If you are a police force, have qualified staff review the results of automatic processing before they are used in acts and measures affecting a person's legal position, and document the review so that it can be traced (art. 3, comma 4).
- If you are a police force, ensure that human oversight of a high-risk AI system is effective and meets Article 14 of Regulation (EU) 2024/1689, with staff who have the necessary competence and training (art. 3, comma 5).
- If you collaborate with the police forces on research or testing of AI for police activity, the collaboration must include clauses excluding the sharing of sensitive operational data and your acquisition or use, even indirect, of AI systems, hardware and software, or AI-integrated devices trained for police activity, except where synthetic data or masked or pseudonymised real data is used (art. 4, comma 2).
- The collaboration agreement must state who owns the intellectual property in the research results, the derived models, the training data and the software, and the police forces always own models trained on sensitive operational data (art. 4, comma 3).
- If you are a police force, run courses on AI applied to police activity in your training institutes (art. 6, comma 1).
- If you are a police force, make AI systems that are under contract, in development, in testing or in use for police purposes compatible with Chapter II of Title I within one year of the decree's entry into force (art. 21, comma 1).
What it makes you log
Logging duty
Article 3, comma 4 requires the human review of automatic results to be documented so that it can be traced, which cannot be shown without a record of the review. It names no log format, content or retention period.
- Kind
- Implicit
- As of
- Provision
- art. 3, comma 4
- Trigger
- automated_decisions
What this law does
Title I of Decreto Legislativo 160/2026 sets rules for the use of AI systems by the Italian police forces, consistently with Regulation (EU) 2024/1689 and Legge 132/2025. Article 1, comma 3 provides that Title I creates no new obligations beyond those the Regulation sets for AI systems and models used for police activities and purposes.
Article 3, comma 4 requires the use of AI systems and models in police activity to provide for qualified human review of the results of automatic processing before they are used in acts and measures affecting the legal position of the persons concerned. The review is carried out by staff identified under each police force's internal procedures and is documented so that it can be traced.
Article 3, comma 5 requires the police forces to ensure that human oversight of a high-risk AI system is effective and meets Article 14 of the Regulation, by staff with the necessary competence and training.
Article 4, comma 2 provides that data processing in collaborations with universities, research bodies and public or private parties includes clauses excluding the sharing of sensitive operational data and any acquisition or use by those parties, even indirect, of AI systems, their hardware and software, or AI-integrated devices trained for police activity, except where synthetic data or masked or pseudonymised real data is used.
Article 4, comma 3 requires the collaboration agreements to state who owns the intellectual property in the research results, the derived models, the training data and the software. The police forces in every case own the models trained on sensitive operational data. Article 5 is the legal basis, under Article 59(2) of the Regulation and Decreto Legislativo 51/2018, for processing of personal data by the police forces within AI regulatory sandboxes where necessary for police purposes.
Article 6 requires the police forces to run courses on AI applied to police activity in their training institutes. Article 21, comma 1 requires AI systems that are under contract, in development, in testing or already in use for police purposes on the decree's entry into force to be made compatible with Chapter II of Title I within one year.
Article 21, comma 2 applies the Regulation's own dates for entry into force and full implementation to the provisions of Title I whose application is provided for by, or depends on, the Regulation.
When LexLint raises it
When your app profile says your app makes high-risk automated decisions or trains models.