
Maciej Szpunar, Advocate General at the Court of Justice of the European Union, published his opinion on Tuesday, November 3 in case C-250/25, which pits the Hungarian publisher Like Company against Google Ireland over Gemini’s answers. It is the first generative AI copyright case to reach that court.
The opinion is not a ruling and does not bind the judges, although they follow such opinions in most cases. A judgment is expected in the coming months.
What the Advocate General said
The opinion separates two things that are often argued together:
- Training the model: using protected articles to train a language model would fall under the EU’s text and data mining exception.
- Answering with those articles: when a chatbot retrieves an article, uses it to build an answer and shows it to the user, the press publisher’s right may be infringed in certain circumstances.
Reports on the opinion point to two risky situations: bypassing a paywall or other technical restriction, or reproducing the wording in a recognizable form. The facts in a news story are not protected, but the way they are told is.
What it did not decide
Szpunar did not say that Gemini infringed Like Company’s rights. That is for the Hungarian courts to settle, using the criteria the European court sets. According to the reports, Google argues that Gemini does not memorize articles and that any problematic output would be covered by legal exceptions.
Why it matters outside Europe
The decision will only apply in the European Union, but Gemini and its rivals work worldwide, and the outcome may change how they cite and display news in the EU. It is also a signal for anyone who publishes content: the debate has moved from “what is AI trained on” to “what does it show the user”.
If your business relies on being found online, this is worth following closely. We wrote a guide on how to show up in AI answers, covering what can be verified today and what cannot.