Spain - Transparency of AI in a work context
Spain has adopted a new legislative provision imposing algorithmic transparency in an employment context. This new provision gives workers the (collective) right to be informed about the parameters, rules and instructions according to which algorithms or artificial intelligence (AI) systems affect their working conditions and/or determine their access to and retention in employment.
What: law
Impact score: 1 – law (legally binding)
For who: policy makers, developers, service providers
URL:
Legislation: https://boe.es/diario_boe/txt.php?id=BOE-A-2021-7840
Information: https://verfassungsblog.de/workers-vs-ai/
Summary
The Spanish statutory provision is the first of its kind and aims to give workers, through algorithmic transparency, a degree of control over the AI systems that impact their work situation. Freely translated, the provision reads as follows:
“The Workers' Council has the right to be informed at regular time periods, by the company, of the parameters, rules and instructions used by the algorithms or artificial intelligence systems, including profiling, that affect decision-making that may have an impact on labor conditions, access to and retention in employment.”
This provision originates from the debate regarding the employment status of meal couriers (self-employed or not). As in several other European countries, Spanish courts have repeatedly ruled that such "riders" are employees, and not self-employed (thus providing them with better social protection). An important factor in that context is that many such digital platforms use algorithms that impose certain minimum requirements. If a courier fails to meet the predetermined requirements, there will be, often automatically, negative consequences (and possibly even a dismissal).
This provision therefore aims to ensure that such algorithms not only respect commercial interests, but also human and labor rights. By imposing a transparency obligation regarding the parameters, rules and instructions used, an attempt is made to bring the functioning of such algorithms out of obscurity.
However, the provision also has its limitations:
- For example, it does not specify how employees can enforce changes to the algorithm or, more generally, what legal actions they could take to enforce their right to information. Thus, there is no real or direct control and the question arises as to how this requirement will in fact be enforced/respected (e.g., how regular is "at regular time periods"? What is understood by "parameters, rules and instructions"?)
- It is not an individual but a collective information right, in favor of an employee council (cf. 'Ondernemingsraad' in Belgium). This means that smaller companies (< 50 employees) do not have to comply with this obligation. In addition, such workers' councils in Spain appear to be bound by confidentiality obligations, which limits the possibility of public debate or investigation.
Finally, the Spanish government plans to create an expert group tasked with examining the effects of algorithms and AI on employment, in an effort to "move toward a fair and rights-based technological transition." This could mean that the provision's shortcomings will be fixed in the near future. In this sense, one of Spain's largest unions has already proposed broadening the scope and terms of the new provision, creating a registry for algorithms, providing a broader right of explanation, a modified liability regime and audit rights - similar to the provisions of the European proposal for an AI act (cf. the requirements that would apply to high-risk AI systems: e.g. transparency, data quality and explainability).