EU court adviser warns Google Gemini could breach publisher copyrights

The legal dispute over news content

An adviser to the Court of Justice of the European Union has outlined potential copyright risks for Google Gemini. Advocate General Maciej Szpunar delivered a nonbinding opinion regarding a dispute between the Hungarian publisher Like Company and Google Ireland. The case explores whether the chatbot violates the rights of publishers when it retrieves and reuses news articles.

The court's explanation focuses on two primary areas of concern. First, it looks at whether the AI was trained on copyrighted material. Second, it examines how Gemini retrieves specific articles to provide answers to user prompts. While the court has not yet made a final ruling, the opinion provides a framework for how these interactions might be judged.

Szpunar suggested that the court should decline to rule on the training questions for now. He noted that the case currently lacks evidence proving that Like Company's specific articles were used to train the Gemini model. However, the ongoing dispute highlights the growing tension between AI developers and the news industry over content ownership.

Defining copyright infringement in AI

The adviser proposed a specific test to determine when an AI's output becomes an infringement. He suggested that a violation occurs if a chatbot reproduces an article's expression so extensively that it acts as a substitute for reading the original piece. This means the AI cannot simply provide a summary; it must not mirror the unique way a publisher wrote their story.

Importantly, the opinion clarifies that reporting the same facts does not meet the test for infringement. AI models are generally permitted to provide factual information as long as they do not copy the specific creative expression of the source. This distinction is critical for understanding where the line is drawn between information sharing and illegal copying.

Additionally, the adviser noted that infringement could arise from unlawful access to content. This includes situations where an AI might bypass effective technical protections to reach a publisher's material. These findings provide a roadmap for how the court might evaluate the behavior of various AI tools in the future.

A landscape of ongoing AI litigation

Google is not alone in facing scrutiny over how its models interact with copyrighted works. The company is currently battling a lawsuit in the United States involving author Scott Turow and several book publishers. These plaintiffs allege that Google copied millions of copyrighted books without authorization to develop Gemini.

Other major tech companies are facing similar legal challenges from prominent media outlets. For example, OpenAI and Microsoft are involved in a major lawsuit from The New York Times regarding the use of news articles for training. Additionally, Amazon has faced a copyright claim from a Twitch streamer who alleged that livestreams were used to train Amazon's AI models.

The legal debate even involves government intervention. In September, the US Justice Department supported OpenAI and Microsoft in their battle against The New York Times. The department argued that restricting the use of copyrighted works for training could give foreign rivals a competitive advantage in the AI space.

Implications for AI procurement

For businesses that build or buy AI tools, this case highlights a vital distinction for compliance. A vendor's explanation of how a model was trained is separate from what happens when a deployed chatbot retrieves external content. Companies must ensure they understand both the origin of the data and the behavior of the live tool.

IT and product teams should conduct thorough reviews of the tools they use. They should ask vendors which specific sources the AI retrieves and how the system handles access restrictions. It is also important to know what safeguards are in place to limit answers that closely reproduce published material.

Procurement teams should also look closely at contractual provisions regarding content rights and responsibility. Testing should go beyond checking for factual accuracy. Teams can compare generated answers against source articles to see if the system is reproducing substantial portions of a publication instead of providing a neutral summary.

What publishers should monitor

Publishers face a different set of challenges as they navigate these legal uncertainties. While traffic loss is a major commercial concern, the legal focus remains on how content is accessed and reused. To build a stronger case, publishers need to provide evidence of how their specific content is being handled by AI systems.

The adviser's opinion does not create a new compliance deadline or require anyone to stop using Gemini immediately. However, it sets the stage for how future disputes will be argued in European courts. It signals that the