An AI chatbot that summarises the news is fine. One that reuses an article so closely that nobody needs to read the original may break European law. That is the line drawn by Maciej Szpunar, First Advocate General at the EU’s Court of Justice, in an opinion published on 3 November in a case against Google.
It is not a ruling. Advocates General advise the court, the judges are not bound, and the judgment will come later with no date set. But this is the first generative AI case to reach that court, so it sets the terms of the argument.
What the case is about
Like Company, a Hungarian publisher of web portals, sued Google in Budapest. It claims Google used its articles to train Gemini and repeatedly reproduced them through the chatbot between June 2023 and May 2024. The Hungarian court then sent four questions to Luxembourg.
Two of them were about training, and Szpunar advises the court not to answer those because they are hypothetical: there is no evidence these articles were used. He adds that training on protected material is, as a rule, allowed under the EU’s text and data mining exception. That part goes Google’s way.

Where he draws the line
The other two questions concern what Gemini does when it looks up a news story in real time and uses it to write an answer. Here, according to the court’s press release, publishers’ rights may be infringed in certain circumstances.
He sets two tests. The first is access: did the chatbot get to the article unlawfully, for example by getting around a paywall or technical measures that block scraping? The second is the answer itself: does it merely report facts, or does it reuse the publisher’s wording so recognisably that the reader has no need to open the original?
Facts stay free. The right protects the investment in the writing, not a monopoly on the news. The chatbot provider is liable if such use is built into its business model, or if it has not put credible measures in place to stop it.
Why Australian businesses should care
The opinion has no force here, but the same fight is under way. In October 2025 the Attorney General ruled out a text and data mining exception, so the training carve out Szpunar relies on does not exist in Australian law. And the ABC and SBS have asked Parliament to bring AI companies into the news bargaining incentive.
If you publish content, the practical point is that how a chatbot uses your work may matter as much as whether it reads it. Our note on Google’s AI Overviews covers another side of how that content now surfaces.
If your business runs its own assistant over web pages or supplier documents, the same test is a sensible design rule: report the facts, link to the source and do not reproduce the text at length. It belongs in your AI policy.
What to watch next
The judges have started deliberating, and they are free to decide differently. Until then, nothing changes for Gemini users. For any company building on this technology, the AI consulting question is simple: if your tool’s answer makes the source unnecessary, ask a lawyer before you ship it.
It will not be the last word either. The IPKat, an intellectual property blog, notes that two more referrals on AI and copyright are already waiting at the same court, so further European decisions are on the way.
