EU court adviser: chatbot answers can infringe publishers' rights
EU Court's first advocate general said chatbot reuse of press articles may infringe publishers' rights.
· Last verified: 5 Nov 2026 (Court of Justice of the European Union (press release 155/26))
Summary
- EU Court's first advocate general said chatbot reuse of press articles may infringe publishers' rights.
- The Hungarian case pits portal operator Like Company against Google Ireland over Gemini's news answers.
- It is the first generative AI case to reach the EU's top court.
The latest
Chatbots that pull press articles in real time and reuse them to generate answers can, in certain circumstances, infringe newspaper publishers' rights, First Advocate General Maciej Szpunar said in an opinion delivered in Luxembourg on 3 November 2026. The Court of Justice described the case, C-250/25, as the first generative AI dispute to reach it. The opinion does not bind the judges.
Details
- The parties: Like Company, which operates a Hungarian online portal, sued Google Ireland, part of the Alphabet group, before the Budapest Regional Court. It argued its articles are protected by the press publishers' right in Article 15(1) of Directive (EU) 2019/790, and that Google used them to train Gemini and reproduced them through it between 13 June 2023 and 31 May 2024.
- The technology: Google Ireland offers Gemini, a chatbot built on large language models that uses retrieval-augmented generation to pull information from the web in real time, according to the Court's statement. It frequently retrieves press articles when answering news questions, the conduct at the centre of the Hungarian claim.
- The referral: The Budapest court stayed proceedings and referred four questions to Luxembourg. Two concern training: whether it falls under the reproduction right or the text and data mining exception in Article 4. The other two concern the generated answers, under the reproduction right and the right of communication to the public.
- On training: Szpunar considered the two training questions inadmissible, citing the absence of evidence that the articles were used in training and the fact that training is not an online use. In the alternative, he said training is in principle permitted under the mining exception, leaving model development outside the publishers' right.
- What the right covers: The opinion states the press publishers' right protects the form of expression, not facts or news as such, and that its purpose is to protect the publisher's investment. That distinction sets the dividing line between an answer that merely relays information and one that reuses protected text.
- The infringement test: Whether the reproduction right is infringed turns on two elements, according to the opinion: unlawful access, such as circumventing a paywall or anti-scraping protection, and whether the answer conveys only facts or reuses the expression to the point that the user no longer needs to visit the original article.
- Communication to the public: The right of making available is in principle infringed where the protected expression is systematically recognisable in answers to related questions, Szpunar said. He did not conclude that every chatbot answer infringes, leaving the assessment to the facts of each case.
- Provider liability: The provider is responsible where the acts are inherent in its business model, or where it fails to put in place appropriate measures to counter them in a reliable and effective manner, the opinion said. That framing ties liability to design choices rather than to individual user prompts.
Background
The press publishers' right was created by the 2019 EU copyright directive to give newspapers leverage over online reuse of their content. Article 4 of the same directive carries the text and data mining exception now invoked for AI model training.
Between the lines
The opinion draws a line through the AI value chain: training is treated as broadly permitted under the mining exception, while the output layer becomes the contested ground. By tying infringement to paywall circumvention and to answers that substitute for the original article, it points publishers toward access controls and substitution evidence as the basis for any claim to payment.
What's next
The judges have begun deliberations, with judgment to follow at a later date. The Budapest Regional Court will then decide the dispute in line with the Court's interpretation, which binds other national courts in similar cases.
Source: Court of Justice of the European Union (press release 155/26)