Live monitoring Updated 72-hour alert levelPriority Editor: Frédéric LENQUETTE
CYBERIACKCybersecurity & AI intelligence · by Frédéric LENQUETTE
RegulationImportant

In Italy, an algorithm can no longer decide alone on a hire or a firing

Italy's decree adapting the AI Act, in force since October 23, reserves the final decision for a human with real authority. Here is what it requires of employers operating in Italy, and what remains allowed.

Translated from French by AI. The original French article is the reference version.

Article illustration : In Italy, an algorithm can no longer decide alone on a hire or a firing
AI-generated illustration

Since October 23, companies in Italy can no longer let software alone decide an employee's fate. Legislative Decree No. 179 of October 7, 2026, published in the Italian Official Gazette on October 8, reserves the final decision on a hire, a sanction or a contract termination for a human being.

The 54-article text adapts Italian law to the European regulation on artificial intelligence (the AI Act) and implements Italy's 2025 AI law (Law No. 132/2025). It does not cover only employment: it also designates the oversight authorities and organizes training for public officials, magistrates, teachers and regulated professions.

What Article 41 forbids employers to do

Article 41 covers decisions to hire, to change or end a contract, disciplinary sanctions and performance reviews. None of them can be made solely by an AI system. According to the text as quoted by the legal site Lavorosì, the final decision is "in any case reserved to a natural person, who exercises effective and autonomous power." In practice, it is not enough for a manager to approve the machine's recommendation with a click: the manager must be able to depart from it.

The decree itself states that a fully automated decision is void, and that the employee's consent is not an exception. Employees are also entitled to clear information about the AI systems used and their weight in decisions affecting them, the law firm DLA Piper points out.

Not everything is banned, however. Automated screening of applications is not among the final decisions: CVs can be processed by AI, provided the GDPR is respected, according to the site InformazioneFiscale.

AI enters workplace risk assessment

Article 42 links AI to Italy's occupational health and safety code. If a system affects work organization, pace or methods, it must appear in the company's risk assessment. Employers must inform and train their employees on the risks specific to these tools.

Who will enforce it?

The decree gives AgID, Italy's digital agency, the role of notifying authority, responsible for the bodies that certify AI systems. The ACN, the national cybersecurity agency, becomes the market surveillance authority and the single point of contact, alongside sector-specific authorities. Fines, enforced by AgID and the market surveillance authorities, can reach €35 million or 7% of worldwide annual turnover for prohibited practices. Professional bodies have six months to adapt their training, and a teacher training plan must start within 90 days, capped at €100 million drawn from the national program "PN Scuola e Competenze 2021-2027."

What to do if you employ staff in Italy

For the employers concerned, the first steps are fairly simple:

  • inventory the HR tools that score, rank or recommend people (recruitment, evaluation, scheduling);
  • appoint, for each type of decision, a person who makes the call, and keep a record of their own assessment;
  • inform employees about the tools used and their role;
  • add these systems to the occupational risk assessment.

In France, Article 22 of the GDPR already protects against decisions based solely on automated processing, but with exceptions, notably explicit consent. Italy closes that door for employment. This national text adds to European requirements that are becoming clearer: the AI Act has just received its first standard. It remains to be seen whether other member states will follow Italy's example.

Method: AI-assisted monitoring, verified sources, article reviewed and approved by Frédéric LENQUETTE. Our editorial policy.

Share this article