Three organizations representing French publishers and authors have taken Meta to court, accusing the American giant of using copyrighted works without authorization to train its generative artificial intelligence models. The Syndicat National de l'Édition (SNE), the Syndicat National des Auteurs et compositeurs (SNAC), and the Société des Gens de Lettres (SGDL) are demanding the removal of the illegally used data and point to a massive copyright violation. According to Vincent Montagne, president of the SNE, Meta has incorporated “numerous works published by its members” into its databases, used in particular to improve its Llama language model. These corpora reportedly include French-language books from the Books3 database, an illegally compiled collection of nearly 200 works, already at the center of legal disputes in the United States.
The plaintiffs denounce a total lack of transparency on the part of Meta and a direct threat to the publishing world. François Peyrony, president of SNAC, is particularly concerned about the emergence of AI-generated texts, which could “compete with real books by authors.” Christophe Hardy, president of SGDL, believes that “the creation of an AI market cannot be to the detriment of the cultural sector” and calls on technology companies to offer compensation to the creators whose works they exploit. For the time being, no estimate of the financial damage caused by these practices has been communicated, but publishers and authors hope that this legal action will encourage major platforms to respect the current legal framework.
This lawsuit is part of a broader trend of opposition between the cultural sector and digital companies, particularly over the issue of copyright and data usage. Last January, Meta admitted to using copyrighted works without authorization in a US legal proceeding, while defending its approach as “fair use.” This dispute in France comes as the European Union is now requiring AI companies to be more transparent about the sources used to train their models. The outcome of this case could therefore have significant repercussions for the regulation of artificial intelligence and the protection of cultural heritage.
