In the first generative AI case to reach the EU's top court, Advocate General Maciej Szpunar has advised that Gemini-style chatbots can infringe press publishers' rights when they fetch articles in real time and serve up recognisable versions of them, while leaving AI training largely untouched.
A small Hungarian web-portal operator has put the business model of AI chatbots in Europe under the spotlight. In an opinion delivered in Luxembourg on 3 November, Advocate General Maciej Szpunar told the Court of Justice of the European Union (CJEU) that when a chatbot such as Google's Gemini retrieves a press article from a publisher's website and reuses it to answer a user, the publisher's EU-law rights "may be infringed in certain circumstances".
The case, C-250/25 Like Company, began at a regional court near Budapest. Like Company, which runs several Hungarian news portals, claims that Google used its articles to train the Gemini models and repeatedly reproduced and made them available through the chatbot between 13 June 2023 and 31 May 2024. The Hungarian court paused the case and sent four questions to the CJEU, according to the court's press release.
Training largely in the clear
Szpunar's opinion splits the dispute in two, and the split favours AI developers on training. He advised the court to declare the two training questions inadmissible because there is no evidence that Like Company's articles were used to train Gemini, and because training is not an "online use" of press publications under the 2019 copyright directive. For completeness, he added that training a model on protected works is, as a rule, freely permitted under the EU's text-and-data-mining exception. The Kluwer Copyright Blog notes that he states bluntly that "the case in the main proceedings has nothing to do with LLM training".
Real-time answers are the risk
The exposure lies in retrieval-augmented generation, the technique Gemini uses to pull live web content into its answers. Szpunar said infringement of the reproduction right depends on two factors: whether the chatbot accessed the article unlawfully, for example by getting around a paywall or anti-scraping measures, and whether its answer merely reports facts or reuses the article's "form of expression" so recognisably that the user no longer needs to read the original. Facts themselves stay free; the right exists to protect publishers' investment, not to hand them a monopoly over the news itself, the court's summary says.
The right of making available to the public is, as a rule, infringed where an article's expression is systematically recognisable in answers to related queries. Crucially for platforms, the operator would be liable where such use is inherent to its business model, or where it has not put in place credible, effective measures to stop it.
Why it matters
Hungarian outlet Globalhír, which reported the opinion on 3 November, stressed the provider-liability point, writing that the chatbot operator "is also liable" if infringement is part of its business model (translated). Brussels Signal called it the first major legal test of generative AI before the CJEU.
The opinion is not binding, although the court often follows its advocates general, and judges are now deliberating. Two further Hungarian referrals on generative AI and copyright, Manchester United Fanatics Klub (C-806/26) and Mobilguru (C-829/26), are pending, according to Kluwer. If the court agrees, European publishers would gain leverage in licensing talks with AI providers, while European AI start-ups building search-style assistants would face a clearer, and potentially costlier, compliance bar. Google has not publicly commented on the opinion in the coverage reviewed.
