By Hamza Faisal

An adviser to the European Union's top court has said that Google's Gemini chatbot may break copyright rules when it pulls in press articles and reuses them to produce answers. The opinion, delivered on Tuesday by Advocate General Maciej Szpunar, is the first generative AI case to reach the European Court of Justice, and it draws a sharp line between training an AI model and what the model says to users.

The case was brought by Hungarian website operator Like Company, which accused Google of using articles from its online news portals. Szpunar's answer was split: using copyrighted material to train a large language model is, he said, freely permitted under the EU's text and data mining exception — and there was no evidence that Like Company's articles had been used to train Gemini at all. The problem, in his view, lies downstream, in the answers the chatbot gives.

What the opinion actually says

Szpunar recommended the court focus its ruling on Gemini's outputs rather than its training data. "When an LLM-driven chatbot retrieves a publication from a press publisher's website, reuses that publication to generate an answer, and provides that answer to the user, the press publisher's right may be infringed in certain circumstances," he wrote.

The qualifier matters. Copyright does not protect the mere facts a news article reports, Szpunar said — only the specific form of expression. A chatbot that restates facts in its own words is on different ground from one that reproduces a publication's expression. Where exactly that boundary falls will be for the court to decide, and the advocate general's opinion, while not binding, is widely treated as a strong signal of the direction the case is taking. A judgment is expected in around six months.

Why this case reaches beyond Hungary

Publishers across Europe have been watching the case because it is the first to put a generative AI chatbot's answers squarely before the EU's highest court. Much of the AI copyright debate so far has centred on training data — whether models may lawfully ingest copyrighted text at all. Szpunar's opinion largely settles that question in Google's favour for publicly available material, then opens a second front: the moment a chatbot serves up an answer built from a publisher's work.

That framing could matter for every AI assistant that retrieves live web content, not just Gemini. If the court follows the opinion, publishers would have a clearer legal basis to challenge answers that lean too heavily on their reporting, and chatbot makers may need to think harder about attribution, licensing, or filtering in how they present sourced material.

Google's response — and what happens next

Google pushed back on the idea that Gemini copies publishers' work. A spokesperson said the chatbot "learns linguistic patterns to create original outputs rather than replicating content, supported by extensive publisher controls that give rightsholders clear choice and the ability to opt out." The company also welcomed what it called the opinion's recognition that EU law permits AI training on publicly available information.

For now, nothing changes: the opinion is advice, not a ruling. But when the court delivers its judgment in around six months, its reasoning on chatbot answers could set the template for how press copyright applies to AI assistants across the bloc — and publishers and AI companies alike will be reading it closely.