Algorithmic transparency in employment: new obligations from the AI Act to the Transparency Decree

1. The facts

   The core issue of algorithmic transparency in Italy is linked to the management of riders. According to reports from publications such as L'Espresso, Avvenire, and Il Giorno, the most significant case involved the Deliveroo platform and its algorithm named Frank.

   The software used reliability and participation scores to assign work shifts, but it was accused of penalizing workers who were absent for legitimate reasons, such as illness or exercising the right to strike. The proceedings concluded with a landmark judgment by the Court of Bologna, which has passed into res judicata, establishing the discriminatory nature of the system.

   Subsequently, journalistic investigations highlighted how, despite reforms, the risk of opaque digital subordination remains. Currently, new proceedings on the merits promoted by organizations like Nidil CGIL aim to obtain the decoding of predictive parameters used by platforms to prevent further indirect discrimination based on algorithms acting as black boxes.

2. The rules in play

   The regulatory framework relies on several key legal sources. Legislative Decree 104/2022 (Transparency Decree), under Art. 1-bis, requires employers to inform workers about the use of automated decision-making or monitoring systems, providing administrative penalties for failure to comply.

   Regulation (EU) 2024/1689 (AI Act) classifies such systems in the workplace as high-risk (Art. 6 and Annex III), imposing strict risk management and human oversight obligations, with penalties reaching up to 7% of global annual turnover.

   On the anti-discrimination front, Art. 25 of Legislative Decree 198/2006 defines indirect discrimination, which occurs when apparently neutral criteria place certain workers at a particular disadvantage. Finally, Art. 22 of the GDPR guarantees the right not to be subject to a decision based solely on automated processing.

3. What the case law says

   Courts of merit have clarified that an algorithm, although technically neutral or blind to personal reasons, proves discriminatory if it cannot distinguish between unjustified absences and absences protected by law.

   Supreme Court jurisprudence has also confirmed that platform workers may benefit from employment protections, making algorithmic transparency essential for verifying the lawful management of the employment relationship.

   In parallel, the Data Protection Authority has sanctioned major platforms for breaching the principles of lawfulness and transparency, finding that the operation of the reputational score was not adequately explained to workers, thereby preventing them from understanding how to improve their working position.

5. Analysis drafted and verified with edit.legal

   To verify the provisions cited in this article, we used edit.legal. Test our legal AI on official sources and apply it to your own matters.

   Try edit.legal AI

8. What it teaches professionals

   First, it is essential to move beyond generic disclosures and draft technical documentation explaining the algorithm's weighting factors. Second, legal practitioners should request the implementation of effective human oversight procedures (human-in-the-loop) to rectify discriminatory automations.

   Third, companies should conduct periodic algorithmic audits to mitigate compliance risks under the AI Act. Fourth, in litigation, transparency does not necessarily mandate disclosing the source code, but requires a clear and thorough explanation of the system's inputs and outputs.

References: D.Lgs. 104/2022 Art. 1-bisRegolamento (UE) 2024/1689 (AI Act)D.Lgs. 198/2006 Art. 25Regolamento (UE) 2016/679 (GDPR) Art. 22

Editorial oversight: Avv. Federico Papa·ICAMUpdated 20 May 2026