Testing the limits of EU copyright: Like Company v Google Ireland

• Introduction

On March 10th, 2026, the CJEU held its first hearing in Like Company v Google Ireland Limited (C-250/25), concerning a preliminary reference made by Hungary on whether responses generated by large language models can reuse journalists’ work. This case marks the first oral hearing of the Court of Justice of the European Union on legal questions specifically addressing generative AI and EU copyright law. The judges will now begin their deliberations. The opinion of the advocate general is scheduled for September 3rd, and a final judgment is expected a few months later. Like Company demonstrates that the underlying issue is not simply how copyright law applies to AI, but rather whether it is suitable for regulating it at all. As was highlighted in the hearing, the doctrines of reproduction and communication to the public presuppose identifiable acts linked to specific protected works, whereas generative AI operates through mechanisms that make those links unclear or indirect. Therefore, the CJEU is effectively being asked either to stretch copyright doctrine beyond its traditional limits or to acknowledge its inadequacy. This judgment will be a decisive moment in determining whether EU copyright can satisfactorily govern AI systems.

• Factual and procedural background

On April 3rd, 2025, the Budapest Környéki Törvényszék in Hungary referred a request for a preliminary ruling to the CJEU, following the procedure enshrined in Art. 267 TFEU. The proceedings concern Like Company, a Hungarian press publisher, and Google Ireland Limited, a subsidiary of Alphabet Inc. Like Company runs several online news portals publishing editorial and audiovisual content, protected under copyright and the press publisher’s right. Google provides access in the European Union to the chatbot Gemini, formerly Bard, based on large language models. LLMs are trained on a large amount of data and are designed for natural language processing tasks. Like Company claims that Gemini reproduced parts of its press articles on user request and that they were also used in the training of the underlying LLM system, without consent or compensation, between June 13th, 2023, and February 7th, 2024. According to the applicant, the chatbot’s responses are acts of unauthorized reproduction and communication to the public, which go beyond the “very short extracts” which are permitted under the Single Market Directive and the rights provided in the InfoSoc Directive on copyright in the information society. Additionally, the use of their work for LLM training also constitutes reproduction outside the limits of lawful text and data mining.

• Questions referred

The two crucial pieces of legislation applicable to this case are the DSM Directive, which strengthened publishers’ rights over their articles, and the InfoSoc Directive, which defines the rules on reproduction and communication of protected works.

The Hungarian court referred the following questions:

• Communication to the public: whether the display, by an LLM-based chatbot, of part of protected press content constitutes an act of communication to the public under the DSM Directive and InfoSoc Directive, and whether the predictive mechanism of the AI system alters that legal qualification.

• Reproduction in training: whether the training of a chatbot using LLM techniques constitutes an act of reproduction under the InfoSoc Directive and DSM Directive.

• Text and Data Mining Exception: if such training is deemed to involve reproduction, whether it qualifies for the exception for text and data mining under the DSM Directive.

• Responsibility for User-Prompted Output: whether the reproduction of protected content by the chatbot in response to a user prompt, which either quotes or refers to the original press publication, can be imputed to the chatbot provider under the same provisions.

These questions illustrate the difficulty of applying traditional copyrights concepts to generative AI systems that do not operate through clearly identifiable acts, making it hard to identify the connection between protected works and generated outputs.

• The hearing

Like Company’s pleading focused not on the function of LLMs, but rather highlighted the broader structural questions raised by the case. The main issue presented is the insufficient protection of press publishers’ rights against AI-driven tools used by platforms to answer users directly instead of sending them to the original news source. Features of chatbots, such as summaries, allow users to access content without visiting the publisher’s website, with resulting implications for their advertising revenue.

Google’s pleading focused on explaining how the technical functioning of its AI models does not violate copyright rules, as the underlying content is already freely available to internet users. Firstly, the modelling of linguistic patterns and the generation of responses are not actions subject to copyright provisions; they use tokenized data patterns probabilistically, which means that similarities with existing articles can occur without the model retrieving or reproducing them. Secondly, publishers have the possibility of opting out of AI uses through Google, which in this case was not done by Like Company.

These diverging positions reflect the tension between traditional copyright law and the challenges of generative AI.

During the hearing, the contrasting opinions of the Member States clearly emerged, particularly regarding the territorial scope of application of EU copyright law. Some states argued for a broader interpretation covering systems commercialized within the EU, even if not trained within the EU. On the other hand, other states warned against this far-reaching approach and argued that copyright law may apply only with concrete acts within that jurisdiction, such as using servers located in the relevant state.

This debate underscores the challenges posed by AI, as the existing framework may prove inadequate to address systems that operate across jurisdictions and through data-driven processes.

• Implications

The Hungarian court’s questions reflect the Member States’ need for clear guidance from the CJEU on AI-related issues. National judges are increasingly faced with cases concerning AI-generated content, and it may be unclear how such content fits within the current copyright framework. At the same time, there is a need to balance the diverging interests of publishers and technology platforms. Publishers argue that these rules were created to stop large digital platforms from benefiting from journalistic content without compensating its creators. They warn that if chatbots can provide answers based on news articles, users may no longer need to visit the original sources, ultimately reducing the advertising and subscription revenues that support journalism. On the other hand, technology companies rely on provisions that permit text and data mining, the automated analysis of large datasets used to train AI systems, unless rightsholders have expressly opted out.

The final decision of the CJEU may take different directions, from dismissing the referral to fully hearing all four questions. In any case, the ruling will be a landmark in European AI copyright jurisprudence.

• Conclusion

Ultimately, Like Company v Google Ireland is not just about whether generative AI infringes copyright, but about whether the current European copyright framework is capable of addressing these issues. The case brings into focus a clear tension at the core of the copyright regulation: rules are built around identifiable acts and specific works, while AI systems content is generated in ways that make those connections harder to trace. In answering the questions referred to, the CJEU will have to decide whether existing concepts can be ‘stretched’ to fit this new reality, or whether they are starting to show their limits. Either way, the implications go beyond this dispute. The judgment will not only shape how EU copyright applies to AI, but also indicate whether the current legal framework is still equipped to respond to the way content is now created and used.

Bibliography

• Like company v google: CJEU holds first-ever hearing on Generative AI and copyright on 10 March 2026 (no date) Like Company v Google CJEU Holds First-Ever Hearing on Generative AI and Copyright on 10 March 2026 - Bird & Bird. Available at: https://www.twobirds.com/en/insights/2026/like-company-v-google-cjeu-holds-first-ever-hearing-on-generative-ai-and-copyright-on-10-march-2026 (Accessed: 02 April 2026).

• Court of justice streaming hearing on whether AI chatbot summarising content from online press publications infringes IP rights of press publishers (2026) EU Law Live. Available at: https://eulawlive.com/court-of-justice-streaming-hearing-on-whether-ai-chatbot-summarising-content-from-online-press-publications-infringes-ip-rights-of-press-publishers/ (Accessed: 02 April 2026).

• News publisher rails against lack of compensation for Google Ai overviews in EU | Courthouse News Service. Available at: https://www.courthousenews.com/news-publisher-rails-against-lack-of-compensation-for-google-ai-overviews-in-eu/ (Accessed: 02 April 2026).

• Hacker, P. (2025) Copyright, AI, and the future of internet search before the CJEU, Verfassungsblog. Available at: https://verfassungsblog.de/copyright-ai-cjeu/ (Accessed: 02 April 2026).

• Court of Justice of the European Union (2025) Request for a preliminary ruling from the Budapest Környéki Törvényszék (Hungary) – Like Company v Google Ireland Limited (Case C-250/25). Available at: https://infocuria.curia.europa.eu/tabs/document?source=document&text=&docid=301042&pageIndex=0&doclang=en&mode=req&dir=&occ=first&part=1&cid=4117231 (Accessed: 02 April 2026).