The first generative AI copyright dispute to reach the EU top court has taken its key procedural step. On November 3, 2026, Advocate General Szpunar delivered a non-binding opinion in case C-250/25, Like Company v Google - a referral from a Hungarian court asking how EU copyright law applies to chatbot answers that build on press articles in real time. For an industry waiting to see how European law treats AI answers about news, this is the first real map of the arguments.
📅 2026-11-04 - Curia (CJEU)
The core of the opinion is a conditional rule rather than a blanket verdict. A chatbot that reuses press articles in real time may infringe Article 15 publisher rights of the EU Copyright Directive in some circumstances - for example when it bypasses paywalls or technical blocks, or reproduces the article so closely that the user does not need to open the original. Two guardrails on that statement matter just as much. First, the opinion is non-binding and the Court of Justice has not decided the case - the referring national court will apply whatever framework the EU court eventually provides. Second, on the most contested issue, the Advocate General urged the Court not to answer the training-related questions, so they were set aside and remain unanswered.
In the specific example examined in the opinion, the Advocate General actually saw likely no infringement. The chatbot answer looked like a hallucinated mix rather than a substitute for the article, and no technical opt-out measures were in place on the source material. Beyond that example, the opinion reads as practical guidance for AI developers: mitigate near-verbatim outputs, respect technical opt-out measures, and mind paywall bypassing in real-time retrieval.
IMPACT for engineers and businesses: teams building retrieval chatbots over news content in the EU now have concrete signposts even before any judgment arrives. Respect opt-outs and technical blocks at the retrieval layer, avoid near-verbatim output in generation, and treat paywall bypass in real-time retrieval as a legal risk rather than an engineering trick. In practice that means adding source-respect checks to RAG pipelines, testing how close answers sit to a single source, and logging where retrieved content came from. For publishers, the case is a signal that Article 15 is a live instrument against unauthorized chatbot reuse.
The closing thought: output design is becoming compliance design. The teams that treat answer construction - how close it stays to a source and whether the user still needs the original - as a first-class product constraint will ship news-adjacent AI features in Europe without redesigning them under court pressure later.
Sources:
- https://curia.europa.eu/site/upload/docs/application/pdf/2026-11/cp260155en.pdf
- https://curia.europa.eu/
Continue the conversation: https://t.me/AIAutomationArchitecture · Bluesky
Top comments (0)