Spain’s cultural sector says “No, gracias” to Royal Decree – Survey shows: half of authors do not want to license works for AI

Spain was pushing another threat for writers in the training of general-purpose AI models. By proposing an extended collective licensing (ECL) regime to train a large language model in all Spanish languages, Spain’s Ministry of Culture opened a public consultation concerning the royal decree that would set up the basis for forcing writers and other authors of book and media works into extended collective licenses for mass exploitation. The public consultation was closed on Dec. 10, 2024, but led to a wave of concerns articulated by writers, translators, journalist and many more, including by the EWC’s member organisations in Spain.

Background: In February 2024, Spanish Prime Minister Pedro Sánchez presented the proposal to develop an open-source large language model trained in all Spanish languages, to provide Spain’s start-ups a “competitive edge in Latin American markets”. Hence, using extended collective licensing (ECL) can significantly undermine authors’ rights to license exclusively, as the Joint Letter by the European Writers’ Council (EWC), Asociación Colegial de Escritores (ACE), Associació d’Escriptors en Llengua Catalana (AELC), and Asociación de Escritoras y Escritores en Lingua Galega (AELG), to the Ministry of Culture, Spain, states:

Now, the decree seeks to allow training the model with copyrighted works introducing extended collective licensing under Art. 12 of the CDSM Directive (EU) 2019/790, under the assumption, as the decree says, that this mechanism allows that Collective Management Organisations grant, given some conditions, non-exclusive authorisations for the exploitation of works and other subject matter on behalf of rightsholders, irrespective of whether or not the rightsholders have authorised the CMO to do so, i.e. in the absence of express authorisation by all of them. The drafted Royal Decree assumes that many rightsholders, given a remuneration scheme, will be happy to cede them: this is a non-proved assertion, and it is a hint of the deep bias the decree shows. 

Indeed, authors as genuine rights owners of any rights for TDM, AI or generative AI purposes, would be required to opt out not only of express reservation rights under the text and data mining exemption (Art 4(3), 2019/790 EU,  but also of extended collective licenses. The Spanish draft royal decree arguments were accusing indirectly writers and further authors,  that their “mass opt-out would generate a significant obstacle to the proper development of artificial intelligence systems that, however, may be strategic for our country.”
Some steps have already been taken with public funding to train MarIA, a large language model in the four official languages of Spain, with collaboration of the National Library.

LETTER OF CONCERN BY EWC, ACE, AELC and AELG – Survey shows: authors don’t want to licence their works for AI training

In late January 2025, three Spanish Members of the European Writers’ Council (EWC), Asociación Colegial de Escritores (ACE), Associació d’Escriptors en Llengua Catalana (AELC), and Asociación de Escritoras y Escritores en Lingua Galega (AELG), issued a letter to Spanish decisionmakers, to remark aspects of the draft decree that are opaque, dangerous and seriously short-sighted and pro-tech biased. 

The letter (English Version to be found here) included important survey findings from an ad hoc survey among 9000 writers, and with the support from Centro Español de Derechos Reprográficos (CEDRO) and in collaboration with other national authors’ associations, to better understand the real opinion of authors when it comes to GAI training with their copyrighted works.

The survey, driven in several national languages and aimed to more than 9.000 writers, show some data that put things in place:

• As much as 96.5% of the more than 800 respondents believe that the use of their work for the training of GAI models required the express authorisation of the author, in line with the ART Principle (Authorisation, Remuneration, Transparency) invented and championed by the EWC. This, as the national associations affirm, is the most secure way to develop GAI models in a fair way and in compliance with Intellectual Property Rights. 

Manuel Rico, President of ACE, points that “Any type of collective license must be absolutely respectful of all the content that is protected, in terms of copyright, by intellectual property legislation. Authorization, Remuneration and Transparency must be guaranteed in all its extremes. That is why we demand a profound modification of the drafted Royal Decree presented by the government of Spain in the GAI model being elaborated. Without authors there is no culture”. 

• Almost half of the respondents (49.9%) would not confer the right even if paid. Out of them, more than half (52 %) argue that the GAI erodes the quality of creative work. More than half (56.2%) of the respondents claim for transparency (how and what for) in the use of data. This is a deal-breaker to effectively consider an authorisation. 

 

 

 

 

 

 

 

As Sebastià Portell, President of AELC states, “A proper and respectful implementation of AI in the Spanish context would mean good news not only for the cultural and professional rights of literary authors, but also for the rights of the citizens related to language diversity, digital rights and the right to access to good, human created, cultural works”. 

• Finally, the survey studies what kind of remuneration (pay per word or per book, for example) should be better. It seems that the payment per word in preferred, but the important question of the amount per word remains unsolved. 

Cesáreo Sánchez, President of AELG, remarks that “The new ways in which Generative Artificial Intelligence technologies use content created by authors in the context of the cultures and languages of the Spanish State, must necessarily be accompanied by respect for the legislation that protects intellectual property. Professional rights are part, with the cultural rights of creators, of the culture that is sustained by each language, which in turn is the clay with which the identity of each people is built. The new digital time cannot be an excuse for creators’ rights to suffer an unjustifiable regression”. 

The draft Royal Decree has been into public consultation between 19 November and 10 December 2024, a really short span of time, and contains other controversial points, as the expected mechanism of express opt-out of the copyrighted works that could be used in GAI training: it proposes, for example, a ten working days window to exercise the opt-out, which is clearly insufficient. Other doubts, like the real representativeness of the Collective Management Organisations capable of granting the rights, are also in doubt, as well as the obligations of the companies benefiting from the granting of those rights. Another important question is that Art. 2 of the draft decree seems to equal TDM and AI training, which is heavily under discussion: AI is not text and data mining (TDM) and it is not covered under the exceptions contemplated in Art. 4 of the CDSM Directive (EU) 2019/790. These and other aspects need to be changed to effectively respect authors’ rights. 

The signatories of the Joint Letter of Concern consider that the projected Royal Decree is unfair, doesn’t balance rights between agents and poses a perilous precedent. The fact that these GAI systems need vast amounts of data to be trained cannot be taken as an excuse to expropriate rights and possessions of authors. The highest good, as a guideline when judging an act or drafting a law, needs to be preceded by the respect of the rule of law.

As EWC President Miguel-Ángel Serrano poses, “The unclear promise (or the hype) of a better and more innovative world due to GAI can’t be compared to the millenary reality of human culture and the existence and importance of books and other forms of written knowledge (as the mere law) that need to be effectively protected by authorities”. To protect that greater good, the signatories plea for a profound revision and modification of the drafted Royal Decree to ensure a scrupulous compliance with authors’ rights and for the opening of a real and respectful debate with the authors, the rightsholders of their texts and the only ones that can grant any authorisation for the use of their legal property. 

Spanish Ministry of Culture steps on the brakes

Spain’s proposal has had the potential to set a precedent for other countries (even outside Europe) and to the detriment of authors’ rights and works. However, shortly after the joint intervention by EWC and its Spanish Members, the Ministry of Culture put a stop on the project: Instead, a “dialogue” shall be initiated with the cultural sector, ‘to find the best way to address the challenges of generative AI and protect the rights of creators,’ the ministry said on January, 28th.  The cultural and authors organisations of Spain demand, at the same time, all illegal artificial intelligence models that already exist on the market shall be withdrawn and that financial compensation be provided for illegal uses that AI companies have made of their works in the past.

Download the Survey findings (pdf, in Spanish).