On April 3, 2025, the Budapest District Court made a request for a preliminary ruling to the Court of Justice of the European Union (“CJEU”) relating to the application of EU copyright rules to outputs generated by large language model (LLM)-based chatbots, specifically Google’s Gemini (formerly Bard), in response to a user prompt. This Case C-250/25 involves a dispute between Like Company, a Hungarian news publisher, and Google Ireland Ltd.

The Hungarian Court asks the CJEU to opine on several questions regarding the interpretation of provisions of Directive (EU) 2019/790, known as the DSM Copyright Directive, and its interaction with certain related provisions of Directive 2001/29/EC, known as the InfoSoc Directive. Specifically, the questions concern the interpretation of Article 15(1) of the DSM Copyright Directive, which grants press publishers the rights of reproduction and making available to the public with respect to the “online use” of their publications by information society service providers (subject to certain exceptions), and Article 4 of the DSM Copyright Directive, which provides an exception to the rights of reproduction and extraction from lawfully-accessible works for the purposes of text and data mining (subject to certain conditions). The referral also seeks interpretive guidance on Articles 2 and 3(2) of the InfoSoc Directive, which provide authors and other creators with the exclusive right to authorize or prohibit the reproduction of their works and the making available of their works to the public.

The parties’ positions in this case can be summarized as follows:

• Like Company claims that Google’s chatbot makes “continual use” of its protected press publications by reproducing and making available those publications without consent, and that this use exceeds the exception for “use of individual words or very short extracts” under Article 15(1) of the DSM Directive, thus harming its economic interests as an online press publisher. Although Like Company accepts that it gave consent for its content to be displayed on the Google search engine, it asserts that this consent does not extend to the communication to the public of that content via the Google chatbot. Additionally, Like Company contends that the training of the LLM underlying the chatbot involved making unauthorized reproductions of its content, beyond what Article 4 of the DSM Copyright Directive permits.

• Google counters that the chatbot’s output did not constitute a reproduction or making available to the public of Like Company’s press publications under Hungarian law or EU law. With regard to the making available claim, Google argues the chatbot does not make protected content available to a “new public” as required by CJEU case law. On the reproduction claim, Google’s arguments are that much of the chatbot’s output comprises facts or content not protected by copyright and that it did not produce identical copies of Like Company’s press publications. Google also asserts that even if the chatbot output were to comprise a reproduction of the underlying press publications, this act is covered by two exceptions to copyright: (1) the exception for making temporary copies (pursuant to Article 5(1) of the InfoSoc Directive), and (2) the text and data mining exception (pursuant to Article 4 of the DSM Copyright Directive). In addition, Google argues that Gemini (Bard) does not have an information database from which it is able to retrieve content at the request of users; instead it relies on the Google Search database to surface modified versions of articles where a user has provided the original version of the article in their instructions (as occurred in this case). Finally, Google asks the Court to consider the balance between the rights of press publishers pursuant to Article 15 of the DSM Copyright Directive and users’ fundamental rights of freedom of expression and information under Article 11 of the EU Charter of Fundamental Rights.

The Hungarian Court submitted four questions that ask, in essence:

1. Does the chatbot’s text output, which closely resembles the content of press publications, comprise a “communication to the public” under the cited provisions. If so, does it matter that the chatbot output is the “result of a process in which the chatbot merely predicts the next word on the basis of observed patterns”?
2. Does the process of training the LLM underlying the chatbot amount to a reproduction under Article 2 of the InfoSoc Directive, given the model’s reliance on observed linguistic patterns?
3. If the answer to question 2 is yes, does the reproduction fall within the text and data mining exception in Article 4 of the DSM Copyright Directive?
4. When a user prompts the chatbot with text matching or referring to the content of a press publication, and the chatbot’s response displays part or all of that content in the output, does this constitute a reproduction made by the chatbot service provider?

This will be the first case in which the CJEU will address legal questions at the intersection of generative AI and EU copyright law. A judgment is expected by late 2026.

The Covington team continues to monitor regulatory developments on AI, and we regularly advise the world’s top technology companies on their most challenging regulatory and compliance issues in the EU and other major markets.  If you have questions about AI regulation, or other tech regulatory matters, we are happy to assist with any queries.