Rome Court: Algorithmic Transparency and Rider Ranking

1. The facts

   The case stems from an application filed by the trade unions Filcams Cgil, Nidil Cgil, and Filt Cgil against the company Deliveroo Italy S.r.l., as reported by publications such as Il Manifesto and LeggiOggi. The procedural stage corresponds to a first-instance order issued following summary proceedings for the suppression of anti-union conduct. The grievance against the platform concerned its failure to provide detailed information on the operation of the algorithm managing the riders, specifically regarding the parameters used for ranking and shift allocation. The unions argued that the system's opacity prevented any verification of whether the algorithm penalized workers exercising fundamental rights, such as striking or sick leave.

   The court found that the company had failed to comply with the statutory information obligations provided for by recent legislation, having omitted to explain the logic underlying automated decisions. The Court therefore declared the behavior anti-union, emphasizing how the lack of transparency generated an information asymmetry that rendered social and union scrutiny of working conditions impossible. The decision ordered the company to cease the unlawful conduct, in addition to supplying the required technical data.

2. 

   The legal framework

   The regulatory core of the dispute rests on Legislative Decree 104/2022, known as the Transparency Decree, which inserted Article 1-bis into Legislative Decree 152/1997. This provision obliges employers to inform workers about the use of automated decision-making systems that affect the employment relationship, specifying the targets and logic of the algorithms.

   1. Article 28 of Law 300/1970 provides the procedural mechanism used by trade unions to seek the cessation of conduct prejudicial to union freedom.
   2. Article 40 of the Equal Opportunities Code (Legislative Decree 198/2006) is relevant for the burden of proof, providing for a lower evidentiary threshold when factual elements support a presumption of discrimination.

3. What the courts say

   Case law has clarified that the right to algorithmic transparency cannot be sacrificed in the name of trade secrecy. Established precedents affirm that although an algorithm represents a protected corporate asset, its practical application in personnel management must be accessible to allow verification of compliance with constitutional rights. On several occasions, courts have held that an algorithm failing to distinguish between unexcused absences and absences resulting from the exercise of a protected right, such as striking, operates unlawful discrimination by applying identical treatment to legally distinct situations. Furthermore, jurisprudence has specified that information must be provided in understandable language rather than through the mere disclosure of indecipherable source code, thereby ensuring the principle of data accessibility.

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

   1. Corporate compliance programs must incorporate preventive algorithmic audits to ensure that ranking parameters do not generate indirect discriminatory effects.
   2. In litigation, counsel must properly handle the rules governing the shift in the burden of proof, preparing robust statistical evidence to demonstrate system opacity.
   3. It is essential to distinguish between source code transparency and the transparency of operational logic, focusing legal arguments on the latter to secure effective judicial remedies.

References: D.Lgs. 104/2022D.Lgs. 152/1997Legge 300/1970D.Lgs. 198/2006

Editorial oversight: Avv. Federico Papa·ICAMUpdated 31 October 2026Content drafted with AI support and subject to editorial source checks. Despite these controls, inaccuracies may remain: reports and rectification requests are welcome. Report a correction