In brief
An EU Court adviser warned that Google's Gemini AI chatbot may violate press publishers' copyright rights by retrieving and reproducing news articles, especially if it bypasses paywalls or reproduces distinctive expression. The opinion was submitted in response to a complaint from Like Company in Hungary, which alleges that Google used its articles to train Gemini. The adviser noted that liability could arise if such practices are inherent to the business model or if adequate safeguards are not implemented.
In a notice presented to the Court of Justice of the European Union on Tuesday, First Advocate General Maciej Szpunar warned that Google’s Gemini AI chatbot could breach press publishers’ copyright‑related rights by retrieving and reusing news articles to generate responses. The warning comes amid a complaint filed by the Hungarian media company Like Company against Google Ireland, alleging that Gemini used Like Company’s articles to train its large language model and routinely reproduced their content for public use.
Like Company claims that Gemini accessed and used its published content to train the model and then repeatedly reproduced it in chatbot responses. The company’s case, heard by the Budapest Regional Court, was referred to the EU court to seek clarification on whether using press articles for training AI models and reproducing that content in chatbot responses violates publishers’ rights under EU law.
Szpunar’s opinion explains that publishers’ rights could be infringed when an AI system fetches an online article and reuses its content to generate an answer. He noted that infringement could arise if the chatbot accessed the article unlawfully by, for example, bypassing paywalls or technical restrictions. Additionally, if the chatbot reproduces the article’s distinctive expression — rather than merely reporting factual content — and does so in a recognizable form that makes it unnecessary for users to view the original publication, this could also constitute infringement.
The adviser added that publishers’ rights to make content available to the public could be infringed when such reproductions occur systematically in responses to related user queries. He argued that liability could attach to chatbot providers if such practices are inherent in their business models or if they fail to implement appropriate technical and organizational measures to prevent infringement effectively.
While the opinion raises the possibility of infringement, it does not yet resolve the legal question. The Court of Justice remains to decide whether the specific actions of Gemini, including its training methods and response generation mechanism, fall within the scope of infringement. The Case also extends to the question of whether the chatbot’s training data acquisition involved bypassing technical safeguards such as paywalls, a factor that courts will scrutinize closely.
Like Company’s complaint emphasizes the economic impact of unauthorized use of its content, arguing that Gemini provides users with ready–made excerpts that replace the need to access paywalled articles. Consequently, it claims that publishers could suffer a loss of revenue and a dilution of control over the distribution of their content. The case has drawn attention from other media groups across the EU concerned about AI training on copyrighted news material.
Google Ireland has not yet provided a formal response to the complaint. The ongoing legal scrutiny represents a broader inquiry into how AI models interact with copyrighted material and the responsibilities of content providers when they offer responses that incorporate user queries about news topics. The EU court’s decision will likely guide the industry on permissible AI training practices and the boundaries of content reproduction within chatbot interfaces.
In summary, the adviser’s opinion highlights that Gemini could potentially infringe publishers’ rights if it accesses content in ways that violate technical restrictions or reproduces protective expressive content in a manner that makes the original unnecessary for users. Whether these concerns translate into legal liability depends on the Court of Justice’s future rulings on the case and on how the chatbot’s training process intersects with EU copyright parameters. The case also remains unresolved on key points, such as the precise nature of the training data acquisition, the extent of the chatbot’s content reproduction, and the adequacy of any protective measures Google has implemented to ensure compliance with EU copyright rules. The court’s forthcoming decision will clarify these uncertainties and set a precedent for the intersection of AI technology and press copyright within the European Union.
At present, the exact legal dynamics of the Gemini–Like Company dispute remain unclear, as the court has yet to rule on whether the chatbot’s methods of sourcing and reproducing news material truly infringe publishers’ rights under EU law. The case also leaves open questions regarding the obligations of AI developers to safeguard against copyrighted content leakage, the definition of “bypassing technical restrictions” in digital systems, and the extent to which training on publicly available articles may or may not constitute infringement. These factors will shape the trajectory of AI‑copyright litigation across the EU in the coming months. The court’s ruling will be closely watched by publishers, tech developers, and policy makers alike for its implications on AI model training and content licensing frameworks.
