What happened
Advocate General Maciej Szpunar issued a non-binding opinion on November 3, 2026, in case C-250/25, stating that Google's Gemini chatbot may infringe the rights of press publishers under EU copyright law. The opinion distinguishes between the training phase, which Szpunar argues falls under the text-and-data-mining (TDM) exception, and the phase, where generating answers that reproduce protected expression from publisher websites may constitute infringement. The case involves Hungarian publisher Like Company suing Google Ireland Ltd. for reproducing substantial parts of its articles without permission.
Advocate General Maciej Szpunar delivered a non-binding opinion on November 3, 2026, in the case of Like Company v. Google Ireland Ltd. (C-250/25). The case was referred to the Court of Justice of the European Union (CJEU) in April 2025, with an oral hearing held on March 10, 2026. Like Company, a Hungarian press publisher, alleges that Google's Gemini chatbot reproduced and communicated substantial parts of its protected press articles without permission.
Szpunar's opinion draws a critical distinction between two stages of AI system operation: training and . He argues that the training phase, where large language models ingest vast amounts of text to learn patterns, is covered by the EU's text-and-data-mining (TDM) exception. However, the inference phase, where the model generates responses to user queries, is where copyright infringement may occur. Specifically, if a chatbot retrieves a publication from a publisher's website, reuses it to generate an answer, and provides that answer to the user, the publisher's rights may be infringed in certain circumstances.
The opinion clarifies that copyright protects the specific expression of facts in a publication, not the underlying facts themselves. This means that while an AI model can learn from the facts in news stories, it cannot reproduce the unique way those facts are expressed by the publisher. The relevant legal provision is Article 15 of the EU's Digital Single Market Directive, which grants press publishers specific rights over online uses of their publications.
The case is notable for being the first CJEU reference to directly address the intersection of AI outputs and press publisher rights. While Advocate General opinions are advisory and the court's judges are not obliged to follow them, this opinion provides a detailed legal analysis that will likely influence how similar cases are argued and decided across the EU.
Source details: cryptobriefing.com
Why it matters
This opinion is significant because it is the first CJEU reference to squarely address AI outputs and press publisher rights, establishing a legal framework that separates the permissibility of training on copyrighted data from the liability of generating infringing content. For AI developers, this creates a dual reality: while training on copyrighted material may be protected by the TDM exception, the risk shifts to the stage, requiring robust mechanisms to prevent models from reproducing protected expression. For publishers, it provides a potential legal basis under Article 15 of the Digital Single Market Directive to challenge AI-generated content that mirrors their work. The final ruling by the CJEU judges, who are not bound by the Advocate General's opinion, will likely set a precedent for how AI copyright disputes are handled across the EU, influencing both the development of AI systems and the business models of news organizations.
The distinction between training and has significant implications for AI developers. The TDM exception provides a degree of legal certainty for the training phase, allowing models to learn from copyrighted material without immediate liability. However, the risk shifts to the inference stage, where developers must ensure that their models do not generate outputs that reproduce protected expression. This may require additional technical measures, such as content filtering or altered training data, to mitigate copyright risks.
For press publishers, the opinion offers a potential legal footing to challenge AI-generated content that mirrors their work. Article 15 of the Digital Single Market Directive provides a basis for publishers to assert their rights over online uses of their publications. If the CJEU follows Szpunar's reasoning, publishers may have a stronger case against AI companies that generate content infringing on their copyrights.
The final ruling in this case will likely set a precedent for how AI copyright disputes are handled in the EU. Given the growing use of AI in content generation and the increasing number of copyright claims against AI companies, the outcome of this case could have far-reaching effects on the AI industry and the publishing sector. It may also influence how other jurisdictions approach similar issues, as the EU often sets standards for digital regulation.
Interactive Mechanism: How It Actually Works
Explore the underlying technology behind this development interactively.
Why can ethical evaluation not be reduced to one model score?
What to watch next
Monitor the final ruling from the Court of Justice of the European Union in case C-250/25, as the judges are not obligated to follow the Advocate General's reasoning. Watch for how AI developers adjust their processes to mitigate copyright risks, such as implementing filters or altering training data usage. Observe the reactions from the publishing industry and other AI companies regarding the TDM exception's applicability to training versus inference. Track subsequent legal cases in the EU that may cite this opinion or the final ruling to establish broader precedents for AI-generated content and intellectual property rights.
The final ruling from the CJEU in case C-250/25 is the most immediate development to watch. The judges are not bound by the Advocate General's opinion, so the final decision could differ in scope or interpretation. The timing of the ruling is not specified in the source, but it is expected to follow the opinion by some margin.
AI developers may adjust their practices in response to the opinion and the eventual ruling. This could include implementing more robust content filtering mechanisms, altering training data to avoid copyrighted material, or developing new models that are less likely to reproduce protected expression. Companies may also seek legal advice to ensure compliance with EU copyright law.
The publishing industry is likely to monitor the case closely, as the outcome could affect their ability to protect their content from AI-generated infringement. Publishers may also explore new business models or partnerships with AI companies to ensure their content is used in a way that respects their rights.
Other legal cases in the EU and beyond may cite this opinion or the final ruling to establish broader precedents for AI-generated content and intellectual property rights. This could lead to a wave of similar cases, as publishers and AI companies navigate the evolving legal landscape.