Italy's AgID Consults on AI Guidelines for the Public Sector

On 12 March 2026, Italy's Agency for Digital Italy (AgID) launched a public consultation on two new sets of guidelines governing the use of artificial intelligence within public administrations. The consultation — which closed on 11 April 2026 — invited comments, proposed amendments, and contributions from all interested parties via the Forum Italia platform.

The guidelines were adopted through Determination No. 43/2026 and are issued pursuant to Article 71 of the Digital Administration Code (CAD). They were anticipated in the 2024–2026 Triennial Plan for ICT in the Public Administration and represent a key milestone in Italy's implementation architecture for the EU AI Act.

These guidelines are significant beyond the Italian context: they represent one of the first national-level attempts by an EU member state to operationalise the AI Act's principles specifically within the public sector's regulatory and procurement cycles.

The two guidelines at a glance

Guidelines on the development of AI Systems in the PA	A strategic tool for designing new AI-based systems. Introduces the concept of AI stack technological layers, autonomy levels, a logical reference architecture, and the use of personas in public administration workflows. Bridges adoption principles with procurement methodology across the full AI lifecycle.
Guidelines on procurement of AI systems in the PA	A methodological and operational framework for acquiring, managing, and monitoring AI solutions. Frames procurement as a strategic lever connecting technological choices to economic, organisational, and contractual dimensions, with a focus on investment sustainability and public accountability.

Analytical observations

From a regulatory quality standpoint, the initiative is notable for its participatory approach: the public consultation is formally embedded in the standard norm-making procedure under the CAD, signalling that AI governance is being treated with the same procedural rigour as other binding digital regulations.

The explicit focus on the full lifecycle — from development and design through to procurement and monitoring — reflects a systemic approach aligned with the EU AI Act's requirements for high-risk AI systems. In particular, the emphasis on autonomy levels and reference architectures suggests an attempt to standardise what has so far been fragmented administrative practice.

The framing of procurement as a "strategic lever" is analytically relevant: it positions contracting authorities not merely as passive consumers of technology, but as active shapers of AI governance through their purchasing decisions. This resonates with broader debates on public procurement as a regulatory instrument in the EU single market.

Notably, the guidelines do not yet appear to explicitly address the environmental dimension of AI systems in public administration — an area of concern central to the SUSTAIN-AI-REG research agenda. Whether sustainability criteria are embedded in the procurement evaluation frameworks is a question warranting further scrutiny once the final text is published.

Source: AgID — Agenzia per l'Italia Digitale, press release of 12 March 2026. The consultation documents are available for download on the AgID website. Public comments were submitted via Forum Italia (channels: LG Sviluppo and LG Procurement).