Implementation of the AI Act: what changes for companies and the labor market
Italy is taking a decisive step in defining its approach to artificial intelligence and, more broadly, to digital transformation.
On June 10, the Council of Ministers preliminarily approved the legislative decrees implementing the EU AI Act and Law 132/2025, placing the country among the first in Europe to adopt a comprehensive and operational regulatory framework on the matter.
The regulatory intervention does not simply transpose European provisions, but introduces a clear and structured vision of the role of artificial intelligence in society. The guiding principle is explicit: AI must remain a tool at the service of humans.
This choice translates into a strongly anthropocentric regulatory framework, in which transparency, accountability, and human oversight become non-negotiable elements.
In this context, the decree has a dual significance.
On one hand, it provides legal certainty to businesses and public administrations that are required to integrate AI into their processes.
On the other hand, it introduces a series of safeguards for workers and citizens, aiming to prevent technological adoption from producing distortive social and economic effects.
Work and Algorithms: The Primacy of Human Decision-Making
One of the key pillars of the measure concerns the relationship between artificial intelligence and the world of work.
The legislator directly addresses a topic already central to international debate: the risk that automated systems may take critical decisions without adequate oversight mechanisms.
The rule establishes that any employment-related decision—including dismissals, downgrading of roles, or contractual changes—taken exclusively on the basis of an AI system is to be considered null and void.
This represents a clear stance that effectively introduces a general obligation of human supervision.
The principle of so-called “human in the loop” thus becomes a legal requirement, no longer merely a best practice.
The objective is to prevent high-impact decisions on people’s lives from being delegated to opaque systems, often difficult to interpret even for their own developers.
This approach also extends to hiring processes.
Companies, including public administrations, may continue to use AI tools for candidate screening, CV analysis, or training activities.
However, the final hiring decision must necessarily be made by a human being.
The rule thus addresses one of the most critical issues of contemporary automation: the growing spread of algorithmic recruitment systems, often accused of replicating or amplifying pre-existing biases.
The obligation of human intervention introduces an element of accountability that could reduce these risks, although not eliminate them entirely.
Workplace Safety in the Age of AI
Another innovative element concerns the extension of occupational safety regulations to the use of artificial intelligence.
The decree requires AI systems to be included in the workplace risk assessment provided under Legislative Decree 81/2008.
This is a significant step, as it broadens the very concept of occupational risk.
It no longer refers only to physical dangers, but also to psychological, organizational, and technological risks.
For example, AI-based performance monitoring systems may generate stress or undue pressure on workers; likewise, algorithmic malfunctions in industrial contexts could have significant operational consequences.
Including AI in risk assessments requires companies to upgrade their safety management practices, demanding interdisciplinary expertise ranging from engineering to occupational psychology.
Fair Compensation and the Value of Intellectual Work
The decree also addresses an issue often overlooked in the AI debate: the economic impact on professionals.
Article 48 introduces the principle of fair compensation for work involving AI systems, aiming to prevent technological efficiency from translating into a generalized reduction in fees.
The logic is twofold.
On one hand, it seeks to counter the possible “commoditization” of intellectual work, where AI use becomes a lever to reduce costs.
On the other hand, it recognizes that integrating these tools introduces new responsibilities and professional risks that must be adequately compensated.
This aspect is particularly relevant for categories such as lawyers, consultants, journalists, and technical professionals, who are increasingly exposed to the integration of automated systems into their workflows.
Intellectual Property and Competitiveness
On the industrial front, the decree introduces a significant measure for protecting innovation.
Algorithms used to train artificial intelligence systems may be protected as trade secrets.
This provision strengthens the position of companies investing in research and development, offering greater legal certainty over the protection of the so-called “technological core” of AI systems.
In a highly competitive global context, where value is increasingly concentrated in models and data, the protection of intangible assets becomes a strategic element.
At the same time, the rule sits within a delicate balance between property protection and transparency, especially in cases where AI systems affect fundamental rights.
Institutional Architecture and Oversight Powers
To ensure effective enforcement of the new rules, the decree defines a structured governance system. The two main authorities are:
The Agency for Digital Italy (AgID), responsible for notification, certification, and technical coordination
The National Cybersecurity Agency (ACN), designated as the market surveillance authority
They will be supported by sector-specific authorities, including the Bank of Italy, Consob, and Ivass, each with competencies in their respective domains.
The system provides significant inspection and sanctioning powers, in line with the EU AI Act framework.
The most serious violations—such as the use of prohibited practices (e.g., emotion recognition systems in the workplace, except in specific cases)—may result in fines of up to €35 million or 7% of global annual turnover.
This sanctioning level is comparable to that of the GDPR, confirming the importance the European legislator assigns to AI regulation.
The European Timeline: A Gradual Transition
National implementation is part of a broader European roadmap, with progressively phased enforcement:
From August 2, 2025: obligations for general-purpose AI models (GPAI), including transparency requirements on training data and copyright compliance
From August 2, 2026: requirements for high-risk AI systems, including those used in workforce management, recruitment, and critical decision-making processes
From August 2, 2027: full application of rules for AI systems embedded as safety components in regulated products
This gradual approach is intended to allow businesses and institutions to adapt progressively, avoiding regulatory shocks and supporting internal skill development.
Beyond Regulation: The Challenge of Skills
Beyond legal aspects, the decree raises a broader issue: the cultural and technical preparedness of the national system.
The introduction of strict rules requires significant investment in digital literacy.
This is not only about using AI tools, but about understanding their limits, risks, and implications.
From data management to bias prevention, from cybersecurity to algorithmic transparency, the required skills are cross-disciplinary and constantly evolving.
For businesses, this means redesigning internal processes and strengthening compliance functions.
For public administrations, it implies developing adequate technical capacity to evaluate and oversee the systems they adopt.
A European Model in the Making
With this intervention, Italy positions itself within the broader European attempt to define a regulatory model distinct from both the United States and China.
A model that aims to combine innovation with the protection of rights, avoiding both deregulation and excessive restriction.
The challenge will be translating principles into practice, ensuring that rules do not become an obstacle to innovation but rather an enabling factor.
This balance represents a crucial test not only for the future of artificial intelligence, but also for the competitiveness and social cohesion of the continent.