# Google’s Gemini chatbot may breach EU copyright laws > Source: > Published: 2026-11-04 04:00:00+00:00 Google’s artificial intelligence chatbot Gemini may infringe European copyright law when it reuses journalistic articles to generate answers for users, an adviser to the European Union’s highest court concluded on November 3. In a landmark [opinion](https://curia.europa.eu/site/jcms/3317819_EmbedVideo/en/opinion-of-the-advocate-general-at-the-court-maciej-szpunar-in-case-c-250/25-like-company) setting out the first major legal test for generative AI before the Court of Justice of the European Union (CJEU), Advocate General Maciej Szpunar argued that while tech giants can, as a rule, freely use protected publications to train large language models (LLMs), serving AI-generated summaries that mimic original journalistic reporting risks breaching press publishers’ neighbouring rights. “When an LLM-driven chatbot retrieves a publication from a press publisher’s website, reuses that publication to generate an answer, and provides that answer to the user, the press publisher’s right may be infringed in certain circumstances,” the court said in its summary of the opinion. The advisory opinion stems from a case brought before Hungarian courts by website operator Like Company against Google Ireland, part of the Alphabet group. The Budapest Regional Court stayed the proceedings in 2025 and referred four questions to the Luxembourg judges. The publisher alleged that its journalistic articles were scraped without consent to train Gemini, called Bard until February 2024, and that the chatbot subsequently reproduced and repurposed its reporting between June 13, 2023 and May 31, 2024. The dispute centres on a July 21, 2023 article on balatonkornyeke.hu, one of Like Company’s portals. It reported Hungarian singer Kozsó’s plan to bring freshwater dolphins to an aquarium at Lake Balaton, western Hungary. Szpunar drew a technical distinction between training an AI model and outputting synthetic responses to web users. Under the EU’s 2019 Digital Single Market Copyright Directive, ingesting copyright-protected material to train machine-learning algorithms is generally permitted under the statutory text-and-data-mining (TDM) exception, the advocate general noted. He nonetheless proposed that the questions on training be declared inadmissible, since there was no evidence the Hungarian publisher’s articles had been used for that purpose. Outputting answers derived from news websites, though, triggers an entirely different legal hurdle. While copyright protections do not cover underlying factual information, Szpunar explained that publishers’ exclusive reproduction and communication rights are, as a rule, infringed if an AI chatbot systematically reproduces the “form of expression” of an article in a recognisable manner. The senior legal adviser warned that infringement occurs when a chatbot’s answer mirrors original journalism “to such an extent that the user, by reading that answer, is entirely dispensed from accessing the original publication.” Another factor, according to the opinion, is whether the chatbot accessed the article unlawfully, for instance by getting round a paywall or measures that block web scraping. The chatbot provider would be liable where such use was inherent to its business model, or where it had failed to take appropriate measures to counter it, Szpunar added. Google had argued that freely accessible articles were implicitly licensed for reuse by chatbots, a position the advocate general rejected. The European Commission, which intervened in the case, urged judges to separate how AI models are trained from what they produce. The opinion lands amid transatlantic lobbying by media companies, who accuse search engine operators and AI developers of cannibalising digital journalism by synthesising news into conversational widgets without paying licensing fees or directing referral traffic to publishers. While opinions from advocates general are non-binding recommendations, the Luxembourg-based court follows their legal reasoning in the vast majority of final judgments. [#AG](https://x.com/hashtag/AG?src=hash&ref_src=twsrc%5Etfw) Szpunar: Opinion in Case C-250/25 on whether a chatbot’s retrieval and reuse of a press publication to generate an answer for a user constitutes “reproduction” and “communication to the public” under [#EUlaw](https://x.com/hashtag/EUlaw?src=hash&ref_src=twsrc%5Etfw).[#CJEU](https://x.com/hashtag/CJEU?src=hash&ref_src=twsrc%5Etfw) [#AI](https://x.com/hashtag/AI?src=hash&ref_src=twsrc%5Etfw) [#Gemini](https://x.com/hashtag/Gemini?src=hash&ref_src=twsrc%5Etfw) [pic.twitter.com/6VStGlB7yY](https://t.co/6VStGlB7yY) — EU Court of Justice (@EUCourtPress) [November 3, 2026](https://x.com/EUCourtPress/status/2117547196136460731?ref_src=twsrc%5Etfw)