One to watch: Like Company v Google

One of the most significant copyright disputes currently before the Court of Justice of the European Union (CJEU), the proceedings in Like Company v Google (Case C-250/25) arise from a claim brought by Hungarian news publisher Like Company against Google in relation to outputs generated by Google’s Gemini chatbot.

The referring court has asked the CJEU to consider how existing EU copyright law applies to generative AI systems, including the use of copyright-protected material in AI training and the extent to which AI-generated outputs may infringe publishers’ rights. This article details the arguments raised at the oral hearing on 10t March 2026. The decision from the CJEU is due to be published imminently.

The case concerns the use of valuable published content by AI systems and the extent to which publishers can control or restrict that use. The underlying issues are of direct interest to academic publishers as AI developers seek access to large quantities of high-quality text and data, while publishers look for ways to protect and derive value from their content.

The case raises three particularly important copyright questions.

First, the court has been asked to consider whether the reproduction of protected content in a chatbot response can amount to a ‘communication to the public’ under Article 3(2) of the Information Society Directive and Article 15(1) of the Digital Single Market Directive. The court has been asked to consider whether the way LLMs work, namely generation of text by calculating the statistically most likely next word based on context, rather than recalling stored content, affects the outcome of this question.

Secondly, the court will consider whether the training of a large language model constitutes an act of reproduction for the purposes of Article 2 of the Information Society Directive. The question is significant because AI systems typically learn from vast quantities of content during training. A finding that training involves acts restricted by copyright could strengthen the position of rights holders seeking greater control over the use of their works in AI development.

Thirdly, if AI training does involve acts of reproduction, the court must consider whether those activities are nevertheless permitted under the text and data mining exception contained in Article 4 of the Digital Single Market Directive. The exception does allow commercial AI training (unless there is an opt out) but the question being asked is also whether the nature of the use by the LLM (namely the systematic use at scale) would still fall within the scope of the exception.

Additionally, the court has also been asked to consider the territorial scope of EU copyright law, and whether it applies where AI training occurs outside the EU but outputs are deployed within it.

Although introduced before the recent surge in generative AI technologies, the exception has become a focal point in debates over AI training. In addition, the court’s interpretation may influence whether certain uses of copyright-protected works for AI training can take place without individual licences, or whether publishers are likely to retain stronger leverage in requiring AI developers to seek permission and pay licence fees for such uses.

The case may also provide guidance on the attribution of AI-generated outputs. One of the concerns raised by rights holders is the possibility that AI systems may generate material that reproduces or closely reflects protected works. While the facts of Like Company concern news articles, the principles articulated by the court could have broader relevance wherever AI systems interact with proprietary content, including scholarly journals, research publications and professional works.

For academic publishers, however, the most significant aspect of the case may be its potential impact on the developing market for AI licensing. Across the publishing industry, attention has increasingly turned towards agreements under which AI developers obtain authorised access to content for training and related purposes. The commercial viability of those arrangements depends, at least in part, on the legal framework governing unauthorised use. Greater clarity from the CJEU could therefore help shape future negotiations between publishers and AI companies, regardless of which side ultimately benefits from the court’s interpretation.

The next milestone will be the opinion of the Advocate General which is due to be published imminently (it was originally set to be published on 3r September). While not binding on the court, Advocate General opinions frequently offer valuable insight into the legal and policy considerations that may shape the eventual judgment. Given the novelty of the issues and the potential economic significance of the outcome, the opinion is likely to be closely examined by publishers, technology companies and rights holders across a wide range of sectors.

Whatever the ultimate result, Like Company v Google is likely to become one of the leading authorities on the application of copyright law to generative AI.

This article was first published in the ALPSP Policy Bulletin, October 2026.