EU court adviser warns Google Gemini may breach copyright rules

The legal challenge to Gemini

An adviser to the Court of Justice of the European Union recently issued a significant opinion regarding Google's Gemini chatbot. The case was brought forward by the Hungarian website operator Like Company, which is challenging how the AI handles press content. The core of the dispute involves whether Gemini breaches copyright rules when it uses articles from publishers to answer user questions.

Advocate General Maciej Szpunar noted that a chatbot might infringe on a publisher's rights in specific circumstances. This occurs when a large language model (LLM)—an AI system trained on vast amounts of text to generate human-like responses—retrieves a publication and reuses it to provide an answer. The adviser emphasized that while facts are not protected by copyright, the specific form of expression used by a publisher is.

The court will deliberate on this matter in the coming months. A ruling against Google could potentially force the company to pay fees for using such content. This would represent a major shift in the current landscape where AI companies often access and process web content with fewer restrictions than traditional search engines.

Training vs. real-time usage

The legal opinion makes a sharp distinction between the training phase of an AI and its active use by consumers. Szpunar stated that using copyrighted material to train LLMs is generally permitted under the text and data mining exception. This exception is a legal provision that allows for the automated processing of large datasets for research and development purposes.

However, the adviser noted that there was no evidence Google used the specific Hungarian publisher's articles to train Gemini. The concern lies in the output phase, where the chatbot provides answers to users. If the chatbot accesses a publication by unlawfully circumventing paywalls or other technical barriers, it may fall under the publisher's right of reproduction.

Google has responded by defending its technology and its adherence to European law. A spokesperson for the company stated that the text and data mining exceptions are vital for the competitiveness of AI development in Europe. They argue that Gemini learns linguistic patterns to create original outputs rather than simply replicating the content it finds on the web.

Impact on news publishers

The dispute highlights a growing tension between big online platforms and the media organizations that provide the underlying information. Publishers have reported that their web traffic is dwindling as AI chatbots increasingly provide direct answers to users. These summaries often replace traditional search engine results, which historically included hyperlinks that directed readers back to the original news sites.

The European Publishers Council has expressed concern over Google's role as a gatekeeper in the search market. They argue that publishers depend on search visibility to reach their audience and survive financially. The council warns that publishers should not have to choose between being visible on search engines and protecting their intellectual property rights from AI services.

The council also noted that Google's control over access to audiences should not become a tool to secure content for AI. They want to ensure that the transition to AI-generated answers does not strip publishers of the ability to monetize their reporting. This case could determine whether publishers will receive a lifeline or face further erosion of their business models.

Google's defense and controls

Google maintains that its AI systems are designed to respect the rights of content creators. The company points out that it provides extensive publisher controls that allow rights holders to make clear choices about their content. These tools are intended to give publishers the ability to opt out of having their material used in certain ways by the AI.

The company's position is that Gemini creates unique outputs based on learned patterns rather than copying text. By focusing on linguistic patterns, the AI aims to provide helpful information without infringing on the specific expression of a news organization. They argue that this approach aligns with existing European laws that permit AI training on publicly available information.

The final ruling from the Luxembourg-based court will be closely watched by the tech industry. Because the court typically follows the majority of its advisers' non-binding opinions, Szpunar's view carries significant weight. The outcome will likely set the standard for how AI companies handle the balance between providing useful summaries and respecting the copyright of the original authors.