EU court adviser: AI chatbots that replace reading the news may breach publisher rights
What's the deal? Advocate General Maciej Szpunar of the Court of Justice of the European Union said on November 3 that AI chatbots can infringe press publishers' rights when they reuse articles to answer user queries. His non-binding opinion came in Like Company v Google, the first generative-AI case to reach the Court. At issue is Google's Gemini chatbot and articles from Like Company, a Hungarian operator of online news portals.
What changes? Szpunar drew a line between summarising facts and copying expression. A chatbot crosses it when the article's wording stays recognisable in the response "to such an extent that reading the article itself becomes pointless," potentially amounting to making the work available to the public.
The specifics: Liability turns on two combined factors — how the content was obtained and how much was reproduced. A breach may occur where a chatbot circumvents an effective technical measure such as a paywall or anti-scraping system. Responsibility would fall on the provider, not the user, where such reuse is inherent to its business model or it fails to take "credible and effective" measures to prevent it.
Why now? A Budapest court referred four questions in March 2025 after Like Company accused Google of using its articles to train Gemini, then reproducing them in responses between June 2023 and May 2024.
What about training? Szpunar urged the Court to declare the two training questions inadmissible, finding no evidence that Like Company's articles trained Gemini. Should the Court examine the point, he said training a large language model on protected works is "as a rule" permitted under the EU's text-and-data-mining exception, provided rights holders have not reserved them.
What's next? The opinion does not bind the judges. A ruling is expected in mid-2027, and two further Hungarian generative-AI referrals are pending before the Court.
The signal: The opinion targets real-time retrieval and outputs rather than model training, landing as the EU's AI Act rules on general-purpose models enter enforcement. For publishers and AI providers, it raises the prospect that substitutive news answers — or access that bypasses restrictions — could create liability even where training itself is lawful.
Read more: curia.europa.eu
Image credit: Affiliate
