CJEU Advocate General delivers first generative-AI Opinion (C-250/25, Like Company v Google Ireland) on training data and chatbot reproduction

On 3 November 2026 Advocate General Szpunar delivered his Opinion in Case C-250/25 (Like Company v Google Ireland), the first generative-AI dispute to reach the Court of Justice. He concluded that training a language model on freely accessible press articles falls, in principle, within the EU text-and-data-mining exception, and that the press publishers' related right under Article 15 of Directive 2019/790 does not reach training. However, a chatbot that fetches an article in real time and reproduces its content so completely that the reader no longer needs to visit the publisher's site may infringe the press publishers' right, and the service provider is liable where that effect is built into its business model or follows from the absence of safeguards. The Opinion is advisory but the Court follows AG opinions in most cases; the judgment is expected around 2027.

This is the first EU authoritative legal map of generative-AI training and retrieval liability. It is favourable to model training (TDM exception applies, no separate publisher right over training) but draws a sharp line at live web-retrieval/reproduction features (RAG, browser-linked assistants) where the provider can be liable for effectively substituting for the source. It will strongly shape the CJEU judgment and EU AI-training copyright practice for years.

AI providers and deployers should (1) confirm training-data sourcing relies on lawful access and is compatible with the TDM exception, (2) review live retrieval/summarisation features that reproduce source content near-verbatim, and (3) implement safeguards (robots.txt/opt-out compliance, avoiding near-complete reproduction that substitutes for the source) to reduce infringement exposure for retrieval-based products.