The Case Like Company v Google Ireland (C‑250/25) marks the Court of Justice of the European Union’s first direct engagement with generative AI and EU copyright law. The preliminary reference, submitted by the Hungarian national court, concerns Google’s Gemini (formerly Bard) product and its alleged use of protected press content published by the Hungarian media company Like Company.
The questions referred to the CJEU for a preliminary ruling are word-for-word as follows:
1. Must Article 15(1) of Directive (EU) 2019/790 of the European Parliament and of the Council of 17 April 2019 [on copyright and related rights in the Digital Single Market and amending Directives 96/9/EC and 2001/29/EC], and Article 3(2) of Directive 2001/29/EC of the European Parliament and of the Council of 22 May 2001 on the harmonisation of certain aspects of copyright and related rights in the information society, be interpreted as meaning that the display, in the responses of an LLM-based chatbot, of a text partially identical to the content of web pages of press publishers, where the length of that text is such that it is already protected under Article 15 of Directive 2019/790, constitutes an instance of communication to the public? If the answer to that question is in the affirmative, does the fact that [the responses in question are] the result of a process in which the chatbot merely predicts the next word on the basis of observed patterns have any relevance?
2. Must Article 15(1) of Directive 2019/790 and Article 2 of Directive 2001/29 be interpreted as meaning that the process of training an LLM-based chatbot constitutes an instance of reproduction, where that LLM is built on the basis of the observation and matching of patterns, making it possible for the model to learn to recognise linguistic patterns?
3. If the answer to the second question referred is in the affirmative, does such reproduction of lawfully accessible works fall within the exception provided for in Article 4 of Directive 2019/790, which ensures free use for the purposes of text and data mining?
4. Must Article 15(1) of Directive 2019/790 and Article 2 of Directive 2001/29 be interpreted as meaning that, where a user gives an LLM-based chatbot an instruction which matches the text contained in a press publication, or which refers to that text, and the chatbot then generates its response based on the instruction given by the user, the fact that, in that response, part or all of the content of a press publication is displayed constitutes an instance of reproduction on the part of the chatbot service provider?
At its core, the dispute raises fundamental questions under the InfoSoc Directive and the DSM Directive: whether the training of a large language model (LLM) constitutes an act of reproduction, whether LLM outputs amount to communication to the public, and how the text and data mining (TDM) exception and the press publishers’ right apply in an AI context. The latter is a noteworthy aspect of the case as the so called press publishers’ right is a neighboring right and has a different and limited scope and duration of protection as well as protection rationale compared to “normal” copyright protection of original works.
The Four Questions – Legally Important, Technically and Factually Shaky?
The national court referred the four questions presented above, addressing (i) communication to the public via output generated by the LLM, (ii) reproduction during LLM training, (iii) the scope of the TDM exception, and (iv) attribution of liability for AI-generated outputs.
Additionally, the question of jurisdiction is interesting in the case, i.e. if the training of the LLM has occurred in the USA, does the relevant EU regulation apply, and if so, on what grounds?
These questions are rightly seen as potentially fundamental for the future interpretation of copyright law in AI context in the EU, and the related practices of companies acting in this intersection.
However, several commentators, as well as an amicus curiae brief submitted in the proceedings, have already pointed out several weaknesses in the reference: the questions appear to conflate different AI technologies and stages of system operation in a confusing manner. In particular, parts of the reference seem to describe mechanisms closer to retrieval‑augmented generation (RAG)—where a system fetches and reproduces external content at query time—rather than the training of an LLM, which involves statistical learning from large data masses without retaining or recalling works as such. Also there seems to be elements in the referred questions more attributable to search engines.
This matters legally. From a copyright perspective, training, fine‑tuning, retrieval, and output generation raise distinct factual and legal issues, yet the questions tend to use
terms such as LLM, chatbot, and AI system interchangeably, without clearly separating these layers. The different technologies also point to different EU legislations.
Why This Matters for the CJEU’s Answer
Preliminary rulings depend heavily on how national courts frame both the facts and the questions. If the reference implicitly assumes that LLM training operates by storing and reproducing works in a manner resembling database retrieval or online searching, the Court may be invited—perhaps unintentionally—to answer legal questions on the basis of an inaccurate and/or limited technical premise.
The risk is not that the CJEU lacks the ability to grasp the technology, but that unclear terminology may narrow or distort the legal analysis. Several interventions at the oral hearing reportedly focused on clarifying how LLMs/the technology actually function and on disentangling training from output‑side uses.
Significance to Businesses
Even with these shortcomings, Like v Google will remain a test case. It is the first time the CJEU is asked to interpret EU copyright law specifically in relation to generative AI systems, and its answers—however cautiously framed—will influence litigation, licensing practices, and AI governance across the EU.
The aforementioned naturally only if the EUCJ does not dismiss or substantially narrow its ruling in the case. In the meantime, the case is definitely one to follow as the forthcoming ruling is expected to have practical consequences for a wide range of businesses, not only AI developers but also media companies, SaaS providers, and enterprises deploying generative AI tools. By clarifying how EU copyright law applies to AI training and outputs, the Court’s decision may directly affect publishing practices, content licensing strategies, compliance costs, risk allocation in AI supply chains, and the viability of optingout or remuneration mechanisms under the DSM Directive, influencing investment and product design decisions also possibly beyond the news/media sector. The ruling is likely to be expected in 2027.
Hannes Kankaanpää
Associate Partner, Counsel, IP & Technology Law, Licensed Legal Counsel