The Court of Justice of the European Union has been asked to rule on a preliminary question submitted by the Budapest Court, which sees, on one side, Like Company, publisher of Hungarian online newspapers, and on the other side, Google Ireland Limited, as provider of GenAI services.
The case: Like Company vs. Google
Like Company accuses Google of reproducing and communicating its articles to the public without consent or remuneration, in violation of its copyright and related rights. The Hungarian publisher claims that Google’s chatbots use journalistic content to generate responses and summaries which, although derived from its own publications, do not necessarily link back to the original websites, thereby reducing traffic and advertising revenue.
Google denies the alleged infringements, arguing that the tokenization and training process of language models does not constitute an actual reproduction. Moreover, the company states that AI services neither store nor disseminate copies of works but generate entirely new text.
Lastly, it contests the applicability of Hungarian law, since the training of the models allegedly did not take place within the national territory.
The preliminary questions
The preliminary questions submitted to the Court concern the coordinated interpretation of Directive (EU) 2019/790 on copyright in the Digital Single Market (DSM Directive) and Directive 2001/29/EC (Infosoc) and, in particular, are the following:
-
whether the display by a chatbot of text identical or similar to content from press publications constitutes a communication to the public within the meaning of Art. 15(1) of Directive 2019/790, and whether the fact that this results from a next-word prediction process is relevant;
-
whether the training process of a chatbot, based on the observation and filtering of language models, constitutes a “reproduction” within the meaning of Art. 15(1) of Directive 2019/790 and Art. 2 of Directive 2001/29;
-
if the answer to the second question is affirmative, whether the reproduction of lawfully accessible works in this way falls within the text and data mining exception provided for by Art. 4 of Directive 2019/790;
-
whether the generation, at the user’s request, of text based on an instruction identical or referring to text from a press publication constitutes a “reproduction” attributable to the chatbot provider.
The points of regulatory friction
Article 15 of the DSM Directive grants press publishers a specific right to remuneration for the online use of their publications, except for “individual words or very short extracts.”
According to Like Company, Google’s activity does not fall under this exception, since the summaries generated by the chatbots are not limited to marginal excerpts. It further claims that Google, through the training and “tokenization” of the chatbot, reproduced the publisher’s works in violation of its rights and that the chatbot’s responses amount to communications to the public.
Google, on the other hand, invokes the need to balance publishers’ rights with the fundamental freedoms of expression and information, enshrined in the Charter of Fundamental Rights of the EU, emphasizing the creative and autonomous nature of GenAI systems. In particular, it argues that the display of chatbot responses does not constitute a making available to the public, as it is directed at the same public already online.
Furthermore, Google stresses that “tokenization” does not involve the storage of copies of data.
Conclusions and perspectives
The dispute highlights the growing tension between editorial protection and technological progress.
The decision that the Court of Justice of the European Union will issue in the coming months carries potentially historic significance for European case law on copyright and GenAI systems.
While in the United States similar disputes have already been the subject of judicial decisions (see Bartz et al. v. Anthropic PBC, No. 3:24-cv-05417), within the European Union such cases are still relatively few.
This gives the forthcoming decision great relevance, as it could shape the development of European jurisprudence on copyright and GenAI systems.
Elena Bandinelli