November 3, 2026, (Inside AI) — An adviser to the European Union's highest court has concluded that Google's AI chatbot Gemini may infringe copyright rules when it retrieves and reuses press publications to generate answers for users. The opinion, issued Tuesday by Advocate General Maciej Szpunar of the Court of Justice of the European Union, stops short of declaring such use illegal outright. It instead sets a conditional standard that could reshape how AI companies build retrieval-based systems across Europe.
The case originated in Hungary, where website operator Like Company sued Google over Gemini's handling of its published content. Szpunar's advisory opinion, while non-binding, carries substantial weight. The Luxembourg-based court's judges follow advocate-general recommendations in the majority of cases, making Tuesday's assessment a likely preview of the final ruling.
The opinion draws a sharp legal line between two stages of AI development: training and deployment. Szpunar found that using copyrighted material to train large language models is "freely permitted under the text and data mining exception." That exception, embedded in the EU's 2019 Digital Single Market Directive, allows researchers and companies to mine copyrighted works for computational analysis. The adviser's reading gives AI developers broad latitude at the training stage, a significant relief for Google and its competitors.
But the retrieval stage is different. When Gemini pulls a press article from a publisher's website and reuses it to compose a response, Szpunar said, "the press publisher's right may be infringed in certain circumstances." The qualifier matters. The adviser did not specify which circumstances trigger infringement, leaving that determination to the court's final judgment.
Read: Meta Deletes AI Scraping Clause After Expert Backlash
Szpunar also clarified the scope of protection. Copyright rules do not cover "mere facts" reported by news outlets, he wrote. Protection extends only to "the specific form of expression." That distinction could allow AI systems to summarize factual content without liability, provided they do not reproduce the publisher's original wording or structure.
Publishers See A Partial Victory
The opinion lands in the middle of a broader transatlantic battle over AI and news content. Publishers across Europe and North America have argued that chatbots built on retrieval-augmented generation effectively republish their journalism without compensation. Google has maintained that its systems operate within existing legal boundaries.
The EU's press publishers' right, introduced in 2019, was designed to give news organizations leverage in negotiations with platforms like Google. Tuesday's opinion suggests that right may extend to AI-generated answers, not just traditional search snippets. That interpretation could force Google and other chatbot operators to license content or risk liability when their systems quote or closely paraphrase press material.
Google did not immediately comment on the opinion. Like Company, the Hungarian plaintiff, has not publicly detailed the specific Gemini outputs it considers infringing.
The case also highlights a structural tension in EU copyright law. The text and data mining exception was drafted before generative AI existed in its current form. Lawmakers envisioned research databases, not chatbots answering consumer queries in real time. Szpunar's opinion attempts to bridge that gap by treating training and retrieval as legally distinct activities, but the boundary between them remains fuzzy in practice.
For AI developers, the practical implication is that retrieval architecture matters. A model that generates answers purely from its trained parameters faces less copyright exposure than one that fetches live web content and incorporates it into responses. That distinction could push companies toward licensing deals with publishers or toward designing systems that avoid verbatim reproduction.
Read: Policy Token Standard Lets Publishers Opt Out of AI Training
The Court of Justice of the European Union will now deliberate. A final ruling typically follows within months, though no timeline has been announced. If the court adopts Szpunar's reasoning, the decision would apply across all 27 EU member states, setting a precedent for how AI chatbots interact with news content globally.
Similar disputes are unfolding elsewhere. In the United States, publishers have sued AI companies over training data. In the United Kingdom, regulators are examining AI and copyright. The EU opinion, while limited to European law, could influence how courts and legislators in other jurisdictions frame the balance between AI innovation and publisher rights.
Sofia spent years studying how machines parse human language before pivoting to cover the systems now rewriting it. She reads technical papers the way others read news and looking for what the numbers are actually saying beneath the headlines.
