On 5 May 2026, Italy's Ministry of Labour and Social Policies issued Decree No. 51, naming the members of four technical-scientific committees — Strategy, Labour-Market Impact, Workers' Rights, and Small and Medium Enterprises — inside the new Observatory on the adoption of AI in the workplace. The Observatory itself was created by ministerial decree signed on 15 December 2025, which in turn operationalised Article 12 of Italy's national AI statute, Law No. 132/2025. With the committees now staffed, the body that is supposed to steer how Italian employers deploy AI finally has a roster.
This is a milestone worth marking precisely because Italy got here first. Law 132/2025 was published in the Gazzetta Ufficiale on 25 September 2025 and entered into force on 10 October 2025, making Italy the first EU member state to enact a comprehensive national AI framework alongside the EU AI Act. The workplace chapter is its most concrete contribution.
The case for a workplace observatory
Start with the strongest version of the argument for this structure. Algorithmic management is not hypothetical. Scheduling systems, productivity scoring, and automated performance flags already shape how millions of Europeans work, often invisibly. Workers frequently cannot see the logic that ranks them, and individual employees have neither the leverage nor the technical fluency to interrogate a model on their own. A standing public body that convenes unions, employers, and technical experts — and that can publish guidance before disputes harden into litigation — is a defensible response. The four-committee design is also sensibly scoped: separating labour-market forecasting from rights enforcement from the specific predicament of SMEs reflects that these are genuinely different problems. And the Observatory sits at the Ministry of Labour, not a generic digital agency, which keeps the focus on employment rather than abstract AI governance.
Article 12 of the law backs this with real transparency duties. Employers must inform staff whenever AI systems are deployed in work processes and communicate the logic and purpose of the system, the data and parameters used, accuracy and cybersecurity metrics, and the mechanisms for human oversight. A worker who knows a model is scoring them, and on what basis, is better placed than one left to guess. That is a proportionate, information-forcing obligation of exactly the kind we support.
Governance is the easy part
The problem is what the 5 May appointments do not resolve. Naming committee members builds the apparatus of oversight without yet producing any of the binding substance. The committees are advisory and study-oriented: their mandate is to monitor, convene, and recommend. The rules with teeth live elsewhere. Under the delegations in Law 132/2025, the government still owes at least one implementing decree defining an "organic framework" for data, algorithms, and AI-training methods, due by October 2026. That decree — not the Observatory's roster — is where penalties, protective measures, and the practical compliance burden on employers will be set.
This sequencing carries a real risk. When a government stands up a visible governance body months before the operative rules exist, it creates pressure for that body to justify its existence by generating activity. An Observatory staffed in May with no binding standard until autumn has every incentive to fill the gap with soft guidance, recommended templates, and best-practice notes that employers — especially cautious ones — will treat as de facto obligations. Italian commentators describe the Observatory's role as developing guidelines and operational standards for a "human-centred and participatory" transition. That is laudable in spirit, but guidance issued ahead of law can quietly become its own compliance regime, unmoored from the proportionality the statute promises and untested against the AI Act's harmonised rules.
The SME committee is the one to watch
The most telling choice in Decree No. 51 is the dedicated SME committee. Italy's economy runs on small firms, and they are precisely the businesses least able to absorb a thick procedural overlay on top of the EU AI Act. A large bank can resource an Article 12 disclosure programme; a 15-person logistics firm adopting an off-the-shelf scheduling tool cannot easily produce impact assessments, accuracy metrics, and union engagement documentation for every system it buys. If the SME committee's work translates into genuine de-risking — model clauses, safe harbours, lightweight templates — it will have earned its place. If instead it becomes a venue for layering bespoke Italian expectations onto firms already navigating EU obligations, it will entrench exactly the fragmentation the single market is meant to avoid.
A fair verdict
Italy deserves credit for taking workplace AI seriously and for grounding its rules in transparency rather than prohibition. Informing workers when they are being assessed by a machine is the right instinct, and a forum that convenes labour and capital before conflict beats reactive litigation. But governance built ahead of substance is a promise, not a result. The honest measure of this framework will not be the May committee appointments — it will be whether the October 2026 implementing decree is narrow, AI-Act-consistent, and genuinely lighter for small firms, or whether the Observatory's interim guidance has already hardened into a parallel rulebook. Italy has built the room. What it puts in it still matters more.