TL;DR
Advocate General Maciej Szpunar advised the EU’s top court on 3 November that Gemini’s use of news articles in its answers may infringe press publishers’ rights, Reuters reports. He also said training models on protected material is “freely permitted under the text and data mining exception”. The opinion is non-binding, but its split between output and training is the part UK publishers and AI firms should read closely.
The case was brought by Hungarian site operator Like Company against Google. Szpunar did not say every chatbot answer infringes. He identified two situations in which an answer engages the publisher’s reproduction right: where the system got to the material by unlawfully getting round paywalls or similar barriers, and where it reuses “the form of expression of that publication in a recognisable manner”. Facts alone stay free. Copyright reaches “only the specific form of expression”, he said.
His view on training favours AI developers, and he said nothing showed the Hungarian publisher’s articles had been used to train Gemini. A judgment is due within months, and Reuters notes that the court follows most such opinions.
Both sides claim something
Google chose to emphasise training. A spokesperson said it acknowledged an opinion “recognising that EU law permits AI training on publicly available information”, and that Gemini creates “original outputs rather than replicating content”.
Publishers focused on search power. Angela Mills Wade, who runs the European Publishers Council, said publishers “should not have to choose between search visibility and protecting their rights”. A ruling against Google could oblige it to pay for such content, Reuters says, which would matter to smaller outlets losing traffic to AI summaries.
Why UK readers should watch
This is a case about EU law, and any UK effect would need its own legal analysis, which Reuters does not offer. In the UK the live pressure on Google has come through competition regulation instead: the Competition and Markets Authority’s conduct requirement on AI Overviews, and a December deadline on AI click data.
In our view, the opinion’s value for UK businesses is as a framework rather than a precedent. It treats how a model learns and what it then says as separate questions, with different answers. Publishers could press the output limb even if training claims fail, and companies building AI search or summarising tools have a clearer reason to test how closely answers track their sources. A final judgment may land differently, so firms should treat this as a signal rather than settled law.