The dispute originated in Hungary, where Like Company sued Google for the unauthorised use of newspaper articles to teach the generative AI chatbot, Gemini, considering that this unauthorised use of its works constituted an infringement of its copyright by reproducing by electronic means the press publications it owned in order to make them available to the public.
Given the novelty of the case and the doubts as to the applicability to that factual situation of the European rules on copyright in the digital single market, the Budapest Court questions whether the extent of the editorial content that can be learned through the answers given by Google Gemini’s chatbot services or the relevance of the data communicated from editorial content that can be learned through the results of a search can reach such a level as to be considered an online use of the publications and, therefore, that such visualisations are subject to both consent and compensation.
Thus, it has asked the CJEU, in summary:
- Whether the display, in the responses of a chatbot based on a Large Language Model (“LLM”), of text which is partially identical to the content of press publishers’ websites to such an extent that it is already protected constitutes a communication to the public; and if so, whether the fact that it is the result of a process in which the chatbot merely predicts the next word on the basis of observed patterns is of any relevance.
- Whether the training process of an LLM-based chatbot that is built on observation and pattern matching, which enables the model to learn to recognise linguistic patterns, constitutes a reproduction.
- Whether such reproduction of legitimately accessible works is covered by the copyright exception that guarantees free use for text and data mining purposes.
- And finally, whether it constitutes a reproduction by the chatbot service provider that, when users give an LLM-based chatbot an instruction that matches the text contained in a press publication, or which refers to that text, the chatbot then generates its response based on the instruction given by the user, a response in which all or part of the content of a press publication is displayed.
It will take a few months, perhaps years, to hear the CJEU’s response, but there is no doubt that the CJEU’s response will shape the future of generative AI in the EU.
Rubén Canales
Lawyer
