Abstract The reference in Like Company v Google (Case C-250/25) is seen as a potential landmark case, giving the EU’s highest court the opportunity to define the scope and conditions of permitted artificial intelligence (AI) training and develop an infringement test for AI outputs. The European Copyright Society (ECS) urges the Court of Justice (sitting as a Grand Chamber) to exercise caution. While the reference stems from a plausible complaint by a press publisher against the provider of an AI powered chatbot reproducing and communicating its editorial content, the implications of this problematic reference could be far-reaching. The reference is factually murky with respect to the technology and services at stake, conflating concepts of “chatbot”, “large language model”, and “search engine”. The reference fails to identify consistently the subject matter at stake, which is the press publishers’ right under Art. 15 of the Copyright in the Digital Single Market Directive (2019/790/EU, hereinafter CDSMD), not authorial works. Specifically, the reference conflates questions relating to the training phase (Questions 2 and 3) with the legal characterisation of the use of press publications by an LLM-based chatbot (Question 4 but also Question 1, referring to the right of communication to the public and the right of reproduction under Directive 2001/29/EC, hereinafter the InfoSoc Directive). If the reference is found admissible, it is suggested that the Court of Justice should address jointly Questions 4 and 1, which relate to the legal characterisation of the use of press publications in the display. Here it is important to correctly understand next-token prediction in large language models, augmented retrieval technology (where the use of data does not generally form part of the learning process) as well as “online use”, defining the scope of the press publishers’ right under Art. 15 of the CDSMD. In the opinion of the ECS, the ambiguous characterisation of a fast-moving technology may result in the failure to realise the societal benefits of AI as a potential general-purpose technology. There are risks that a rash decision will push Europe towards a licensing economy in which AI systems are offered as a service by (non-European) multinationals, without solving issues of equity such as creator consent and distribution of revenues.
Mezei et al. (Mon,) studied this question.