Google’s Gemini chatbot may infringe EU publishers’ rights when it retrieves and reuses news articles in certain circumstances, an adviser to the bloc’s highest court said on Nov. 3.
Advocate General Maciej Szpunar delivered the nonbinding opinion in a dispute between Hungarian publisher Like Company and Google Ireland. The Court of Justice of the European Union has not yet ruled.
For businesses building or buying AI tools that retrieve online content, the case raises questions about how those tools access source material and reproduce it in their answers.
According to the court’s explanation of the opinion, the case concerns both alleged training use and Gemini’s retrieval of articles to answer questions.
Szpunar proposed that infringement could arise from unlawful access, including bypassing effective technical protections, or from answers that recognizably reproduce an article’s expression so extensively that they substitute for reading it. Reporting the same facts alone would not meet that test.
He recommended that the court decline to answer the training questions because the case lacked evidence that Like Company’s articles had been used to train Gemini. His preliminary assessment also suggested that the publisher’s specific chatbot example did not establish infringement, subject to the Hungarian court’s findings.
AI companies running into copyright barriers
Google wouldn’t be the first AI company to come under scrutiny over copyright, with several cases either settled or ongoing involving AI companies and copyright holders. Amazon was hit with a copyright claim by Twitch streamer Warren Pandiscia, who alleged that their livestreams were used to train Amazon’s AI model.
OpenAI and Microsoft were also hit with a major lawsuit from The New York Times, which claims that the companies unlawfully used its news articles to train their models. Google is also battling a court case in the US involving bestselling author Scott Turow and several book publishers, who allege that Google copied millions of copyrighted books and other works without authorization to develop Gemini.
The US government has also intervened in the debate over AI training and copyright. In September, the Justice Department backed OpenAI and Microsoft in their legal battle against The New York Times, arguing that restricting the use of copyrighted works to train AI models could give foreign rivals an advantage.
Separate EU disputes concern competition and search policies
Google has also faced EU competition proceedings, which concern different legal issues from the publishers’ rights dispute. One involved its €4.1 billion (approximately US$4.6 billion) Android case, for which it lost its final appeal earlier this year. The bloc accused Google of using its market dominance to pre-install and favor its own apps over rivals.
Later in July, the company wasfined €890 million (approximately US$995 million) for favoring its own search services and restricting developers from offering cheaper payment and subscription options outside the Google Play Store.
In August, Google also changed its enforcement of the site reputation abuse policy in the European Economic Area following EU scrutiny. The dispute concerned whether penalties for hosting commercial partner content unfairly harmed publishers.
More must-read AI coverage
- SS&C Intralinks DealCentre AI vs. Datasite: Which platform is built for the future of dealmaking?
- SS&C Intralinks FundCentre AI vs. Juniper Square: Which platform better supports modern private markets fund managers?
- Why Data, Not Models, Determines AI Success
- The Rise of the AI-Native Factory: How Physical AI Is Transforming Manufacturing
What businesses using web-connected AI should check
For IT and product teams, the case highlights a distinction worth examining during AI reviews: how a model was trained and how a deployed chatbot retrieves external content are separate questions. A vendor’s explanation of training practices does not necessarily explain what happens when its tool accesses a news website to answer a user.
Teams evaluating these tools should ask vendors which sources they retrieve, how they handle access restrictions, and what safeguards limit answers that closely reproduce published material. Procurement reviews should also examine contractual provisions covering content rights, complaints, and responsibility for disputed outputs.
Testing should look beyond factual accuracy. Teams can compare generated answers with source articles, check whether links support the claims made, and examine how the system responds to requests to reproduce substantial portions of a publication.
Publishers have a different task: preserve examples of disputed outputs alongside the original articles, prompts, and access conditions. Traffic losses may explain the commercial concern, but evidence of how content was accessed and reused is more relevant to the proposed legal test.
These are practical review steps while the case proceeds. The adviser’s opinion does not itself introduce a new compliance deadline or require businesses to stop using Gemini.
Read more: Explore how Google’s changes to its European search spam policy affect publishers hosting commercial partner content, another dispute over publishers’ relationship with the platform.