Italy's civil servants are experimenting with artificial intelligence at a pace that outstrips official policy, according to findings unveiled at Forum PA 2026 in Rome, exposing a widening gap between grassroots adoption and institutional oversight.
Two-thirds of public sector employees now use AI tools in their daily work, yet 59% operate without internal guidelines, formal training, or secure platforms. The data, drawn from a research study titled "La Pubblica Amministrazione infrastruttura strategica del Paese" and presented by FPA (Gruppo Digital360) at the Convention Center "La Nuvola," underscores a troubling paradox: enthusiasm without guardrails.
Why This Matters for Residents
If you interact with Italian public services, this matters directly to you. AI is already shaping how your documents are processed, how benefits are assessed, and how information is provided—often without your knowledge.
• Unregulated usage: The majority of civil servants deploying AI receive no official guidance or structured support from their administrations.
• Basic applications dominate: Document synthesis and text drafting account for most use cases, while advanced automation remains rare.
• Governance vacuum: Only 41% of respondents report receiving any direction from their employer on appropriate AI usage.
• Timing: The survey arrives as the EU AI Act enters full force and Italy's own Law 132/2025 on artificial intelligence takes effect.
What This Means for Your Interactions with Public Services
While the full scope of AI deployment across Italy's public sector remains unclear, residents are most likely to encounter algorithmic decisions in high-volume administrative processes. Services at INPS (the national social security institute) for benefit eligibility screening, Anagrafe (municipal registry offices) for document processing, and Agenzia delle Entrate (tax authority) for audit selection and compliance review all handle thousands of cases daily. These are precisely the areas where efficiency gains from AI are greatest—and where errors or biases could affect you directly.
When you submit an application or request through these channels, you should be aware that an algorithm may initially assess, filter, or prioritize your case. If a decision affects you negatively, you have the right to request human review, and emerging AI legislation now establishes that right more formally. Ask the civil servant directly: "Has AI been used to assess my application?" A straightforward answer should be part of transparent administration.
Additionally, if a decision seems inconsistent with established rules or lacks clear justification, documenting your concerns and requesting escalation to a supervisor creates a paper trail and signals the need for oversight.
What Civil Servants Actually Do with AI
The study reveals a pragmatic, task-oriented approach to generative AI among Italian public employees. The most common applications include document summarization and analysis (59%), text drafting (59%), online information searches (57%), and regulatory research (54%).
Further down the list are language support (36%), data analysis and reporting (36%), and brainstorming assistance (30%). Advanced office automation—the kind that could fundamentally restructure workflows—accounts for just 14% of usage.
This pattern suggests civil servants are treating AI as a personal productivity aid rather than a strategic institutional resource. The tools they choose are often consumer-grade platforms accessed individually, raising questions about data security, compliance, and algorithmic bias.
The Governance Gap
The fact that nearly six in ten employees use AI without formal protocols exposes a critical lag between technological adoption and institutional readiness. The Agenzia per l'Italia Digitale (AgID), the government's digital watchdog, has issued guidelines on AI procurement, development, and deployment in the public sector, covering governance, risk management, transparency, and accountability.
Yet the survey data suggests those guidelines have not translated into day-to-day practice. Only 41% of civil servants say their agency has provided any indication or support regarding AI usage. This leaves a majority navigating ethical, legal, and operational terrain on their own.
The stakes are high. Public sector AI use in Italy now encompasses at least 157 documented projects, from the Agenzia delle Dogane's "Autentica" app, which uses computer vision to detect counterfeit goods, to predictive algorithms for tax policy planning. But without robust internal controls, even well-intentioned tools risk replicating historical biases, violating privacy norms, or eroding due process.
Public Administration Regains Central Role
Beyond the AI findings, the FPA research highlights a broader shift in public sentiment. 53% of Italians now view public administration as a stable reference point for citizens and businesses, considering it essential to economic and social stability. This marks a notable rebound in institutional confidence.
When asked about policy priorities, respondents emphasized price control and inflation management first, followed by employment and job security, and energy security. The list reflects the cost-of-living pressures and geopolitical uncertainty that have defined the mid-2020s in Italy and across Europe.
Forum PA 2026: A Crossroads Moment
Forum PA 2026 brought together policymakers, technologists, and administrators under the theme "Per una PA che genera il futuro" (For a Public Administration that Generates the Future), serving as a focal point for debates on digital transformation, AI governance, and anticipatory policymaking.
The Ministry of Economy and Finance (MEF) and the Department of the Treasury showcased their efforts to build foresight offices and AI governance units. Regional governments, including Abruzzo, demonstrated AI-powered civic assistants like "Iride." And municipalities were measured on their digital maturity through an index developed by FPA for Deda Next, tracking progress toward PNRR (National Recovery and Resilience Plan) digitalization targets.
AgID hosted a session on the governance of AI adoption, framing implementation as a journey requiring clear operational use cases, not just aspirational policy documents. The message: Italy's challenge is less about innovation and more about implementation discipline.
Risks and Red Flags
Experts and unions have flagged multiple concerns about AI's role in public services. Chief among them: job displacement, algorithmic bias, and the erosion of human judgment.
Forecasts suggest AI could automate up to 30% of current job tasks by the end of 2026, with call centers and routine administrative roles most vulnerable. Labor groups are demanding income protection measures and retraining programs to cushion the transition.
Equally troubling is the potential for algorithmic discrimination. If AI systems are trained on datasets that reflect historical inequalities—whether in welfare access, hiring, or law enforcement—they risk embedding those biases into official decision-making. The Fundamental Rights Impact Assessment (FRIA), mandated under the AI Act for high-risk systems, is designed to catch such issues, but implementation remains uneven.
Unions also warn of intensified surveillance and work pressure through algorithmic management, particularly in sectors like healthcare and education. Transparency in how AI monitors or evaluates employees is often lacking, and workers report little say in how these tools are deployed.
The Path Forward
Gianni Dominici, CEO of FPA, argues that transformation in Italy's public sector is "not only digital, but organizational, cultural, and cognitive." He calls for a shift in narrative, urging policymakers and media to highlight success stories not to obscure problems, but to demonstrate that change is achievable.
The data from Forum PA 2026 suggests Italy is caught between two realities: a civil service eager to experiment with AI and an institutional framework still catching up. Bridging that gap will require more than guidelines—it demands training, accountability mechanisms, secure infrastructure, and a cultural shift toward responsible innovation. The sustainability of this adoption pattern will depend on whether institutional frameworks can match the pace of ground-level implementation already underway.
