Readers increasingly use AI chatbots to access news and information rather than visiting newspaper websites. By retrieving online articles and using them to generate answers, chatbots raise questions about the extent to which this use may infringe press publishers' rights.
In Case C-250/25, Like Company, Advocate General Maciej Szpunar examines this question in his Opinion of 3 November 2026, considering how Article 15(1) of Directive (EU) 2019/790 (the Copyright in the Digital Single Market Directive) applies to the use of online press publications by AI chatbots.
Like Company, a Hungarian operator of several online portals, brought proceedings against Google Ireland Ltd, alleging that Google had used its articles through its generative AI chatbot, Google Gemini, without authorisation.
Gemini combines Large Language Models (LLMs) with Retrieval-Augmented Generation (RAG), allowing it to retrieve information from the internet in real-time and generate responses.
Like Company alleged that its articles had been used both to train Gemini's AI models and to generate chatbot responses, without their authorisation. The Hungarian court referred questions to the CJEU on whether these activities could infringe the rights granted to press publishers.
The Advocate General (AG) distinguishes between using publications to train an AI model and retrieving publications to generate responses to users.
1. AI Training
The AG proposes that the questions concerning the use of press publications to train an LLM be declared inadmissible, as there was no evidence that Like Company's articles had actually been used to train Gemini. He also considers that LLM training does not constitute an ‘online use’ of a press publication under Article 15(1), as it takes place during model development rather than as part of an online service provided to users.
Nevertheless, he addresses the issue for completeness, noting that the use of protected material to train an LLM may, in principle, fall within the text and data mining exception under Article 4 of the Directive, subject to its conditions.
2. Real-time retrieval through RAG
The position differs for RAG, which allows a chatbot to retrieve online publications in response to a user's query and use them to generate answers. According to the AG, this constitutes an online use of a press publication under Article 15(1).
The retrieval process may also involve technical reproductions falling within the reproduction right under Article 2 of Directive 2001/29/EC (the Information Society Directive). These may benefit from the temporary reproduction exception, provided its conditions are met, including that the final use of the retrieved material is lawful.
Article 15(1) grants press publishers rights of reproduction and making available to the public in respect of the online use of their publications by information society service providers. These rights protect the particular form in which information is expressed, rather than the facts or information reported in an article.
In assessing whether a chatbot's response infringes the reproduction right, the AG considers whether the publication was accessed unlawfully, particularly through the circumvention of technical measures such as paywalls or restrictions on web scraping, and whether the response reproduces its protected expression to such an extent that users no longer need to consult the original.
Verbatim copying is not required, as a chatbot may rephrase an article and still reproduce protected expression where its original form remains recognisable. Conversely, communicating the same facts in a different form does not, on that basis alone, infringe publishers' rights. Individual words and very short extracts are also excluded from the protection afforded.
The AG further distinguishes the reproduction right from the right to make a publication available to the public. The latter may be infringed where recognisable expression from a publication is systematically reproduced in chatbot responses to related user queries, allowing users to access the substance of the original publication without visiting the publisher's website.
The AG considers that responsibility for copyright-related acts carried out through a chatbot may extend beyond the user who initiates a request. A provider may also be liable where it deliberately intervenes to facilitate the relevant acts, where these acts are inherent in its business model, or where it fails to implement appropriate safeguards expected of a reasonably diligent operator.
These safeguards may include programming RAG chatbots to respect publishers' technical restrictions and using output filters to prevent responses from reproducing protected expression. Where appropriate measures are in place, occasional instances of reproduction should not automatically result in provider liability.
The Opinion also addresses whether publishers implicitly consent to the use of their publications simply by leaving them freely accessible online. The AG rejects the suggestion that the absence of technical restrictions automatically amounts to consent. Instead, implied consent requires that the publisher has been actually and individually informed of the intended use and given simple means of preventing it.
This is particularly relevant because measures designed to block web scraping may also affect a publisher's visibility in search engine results, especially where the same bots are used for indexing and chatbot retrieval. A publisher's decision not to block access should therefore not automatically be interpreted as consent to the use of its publications by AI chatbots. The AG's Opinion offers useful guidance but is non-binding. It remains to be seen whether the CJEU will follow his approach.
Authors: Dr Terence Cassar and Dr Mattea Pullicino
For more information on artificial intelligence and intellectual property, contact us at info@gtg.com.mt.