Google’s Gemini AI chatbot could land the company in legal hot water after a top European court adviser ruled that publishers may hold copyright claims over how the tool repurposes news articles to answer user questions.
The opinion arrives as news organizations across Europe escalate their battle with major technology platforms. Publishers argue that AI chatbots lift journalistic content without payment while siphoning away web traffic that once flowed through search engine links.
Advocate General Maciej Szpunar, advising the Luxembourg-based Court of Justice of the European Union, weighed in on a complaint filed by Hungarian website operator Like Company against the US tech giant.
A final ruling against Google would force the company to compensate publishers for content its AI system reproduces. Such an outcome could offer a crucial revenue stream for smaller online media outlets that have watched audiences shrink as AI summaries replace conventional search results.
Szpunar stated that when a chatbot retrieves a press publication, reuses it to formulate a response, and delivers that answer to a user, the publisher’s rights “may be infringed in certain circumstances.” He specified that Gemini’s outputs fall under reproduction rights when the system bypasses paywalls or other technical protections, or when it reuses “the form of expression of that publication in a recognisable manner.”
Copyright law shields only the specific expression of content, not the underlying facts reported, he added.
Meanwhile, Szpunar affirmed that using copyrighted material to train large language models remains “freely permitted under the text and data mining exception.” This distinction could shape how regulators treat AI development versus AI deployment.
The court typically follows such non-binding opinions and will issue its ruling in the coming months.
Google defended its position, stating that European text and data mining exceptions support AI innovation and competitiveness. A spokesperson said Gemini generates original outputs by learning linguistic patterns, backed by publisher controls that give rightsholders opt-out options.
