LUXEMBOURG — Google's Gemini chatbot may infringe the copyright of press publishers when it reuses their stories to answer users, an adviser to the European Union's top court said on Tuesday, in an opinion that could reshape how artificial intelligence systems handle news content across Europe. Advocate General Maciej Szpunar told the Court of Justice of the European Union (CJEU) that a chatbot which retrieves a press publication, reuses it to generate an answer and delivers that answer to a user can infringe the publisher's rights "in certain circumstances" — most clearly when it circumvents a paywall or reproduces the publication's form of expression in a recognisable way.

The opinion is not binding, and the court's judges are not obliged to follow it. If the court agrees, Google could face pressure to pay licensing fees to press publishers, and other chatbot operators could face the same question about how their systems summarise and quote the news.

Key Points

• Advocate General Maciej Szpunar said Google Gemini may infringe press publishers' rights when its answers reuse their stories.
• Infringement arises when a chatbot circumvents paywalls or technical barriers, or reuses a publication's form of expression recognisably.
• Mere facts are not protected — only the specific form of expression in which they are presented.
• Using copyrighted material to train large language models is freely permitted under the text and data mining exception, Szpunar said.
• The case, C-250/25, was brought by Hungarian website operator Like Company over Gemini's alleged reproduction of its articles between June 2023 and February 2024.

What Happened

Szpunar delivered his opinion in Luxembourg on Tuesday, November 3, 2026, in a case referred by a court in Budapest. Like Company, a Hungarian website operator, alleges that Google's Gemini chatbot reproduced substantial parts of its press articles without permission over an eight-month period, from June 2023 to February 2024. The Budapest court referred the dispute to the CJEU in April 2025, asking how EU copyright rules apply when a large language model chatbot answers questions by drawing on press content.

The core of Szpunar's answer draws a line between two very different uses of copyrighted material. Training a model on published text, he said, is "freely permitted under the text and data mining exception" in EU law. Answering a user by retrieving a specific publication and reusing it, however, is a different act — one that can engage the press publisher's neighbouring right, a protection created to give news publishers a say over the online use of their publications.

Background/Context

The dispute sits inside a much larger fight between news publishers and technology platforms over who gets paid when journalism is reused online. European publishers have spent years arguing that search engines, aggregators and social platforms profit from their reporting while the newsrooms that produced it struggle. The EU's 2019 copyright directive answered part of that complaint by creating a neighbouring right for press publishers, giving them a right over the online reproduction and communication of their publications, subject to exceptions for very short extracts and private use.

Google has already been through that fight in its search business, striking licensing deals with publishers in several European countries after regulators and courts pushed it to pay. The Gemini case asks whether the same logic extends to a new kind of product: not a list of links, but a conversational answer that may itself contain the substance of a story, reducing the need for the user ever to visit the publisher's site.

Google Gemini Copyright: What the Opinion Actually Says

The opinion's most quoted sentence sets out the test in full: "When an LLM-driven chatbot retrieves a publication from a press publisher's website, reuses that publication to generate an answer, and provides that answer to the user, the press publisher's right may be infringed in certain circumstances." The words "in certain circumstances" do most of the work. Szpunar does not say every chatbot answer drawn from a news article infringes copyright. He identifies two routes to infringement: unlawfully circumventing paywalls or other technical barriers, and reusing "the form of expression of that publication in a recognisable manner".

The first route is the more straightforward. A paywall is a publisher's chosen business model and a technical measure; a system that defeats it to extract the text behind it is not merely reading public information. The second route is subtler: how close must an answer be to the original article before it counts as a recognisable reuse of its expression, rather than a lawful summary of unprotected facts? That question will shape how chatbot makers design retrieval, paraphrasing and citation.

Reactions/Quotes

Szpunar's own words set the terms of the debate: "When an LLM-driven chatbot retrieves a publication from a press publisher's website, reuses that publication to generate an answer, and provides that answer to the user, the press publisher's right may be infringed in certain circumstances." Publishers' groups are expected to welcome that recognition, while technology companies are likely to stress that the opinion is non-binding and that facts remain unprotected.

Why It Matters

For readers, the case is about whether the news they get from a chatbot is built on journalism that someone is paid to produce. If publishers cannot monetise chatbot reuse, newsrooms lose revenue; if chatbot makers must license everything they summarise, the cost and complexity of AI answers rise, and some features may narrow or disappear. The opinion's middle path — facts free, expression protected, paywalls respected — is an attempt to keep both the tool and the journalism viable.

What Happens Next

The CJEU's judges will now consider Szpunar's opinion and deliver a final judgment, expected in around six months. Advocate-general opinions are usually, though not always, followed. If the court adopts his reasoning, the case returns to the Budapest court to apply the ruling to Like Company's specific claims about Gemini's output between June 2023 and February 2024. Either way, publishers and AI companies across Europe will be reading the final judgment closely, because it will set the terms for the next round of licensing negotiations — or the next round of litigation.

Frequently Asked Questions

Is the opinion binding on Google? No. An advocate general's opinion advises the court; the judges make the final, binding ruling, expected in about six months.

Does this mean chatbots cannot use news at all? No. Szpunar said mere facts are not protected, and training models is permitted under the text and data mining exception.

What is case C-250/25? It is the CJEU case brought by Hungarian operator Like Company against Google, referred by a Budapest court in April 2025, over alleged reproduction of its articles by Gemini between June 2023 and February 2024.

Could Google have to pay publishers? Possibly. A ruling against Google could force licensing fees for chatbot use of press content, as already struck for search in parts of Europe.

Does the ruling affect users in the US? Not directly, but global AI products are often designed to a single standard, so a European ruling could change how chatbots handle news everywhere.

Related coverage: Pentagon–Anthropic talks put Claude military limits to a Friday test and Apple earnings beat forecasts as AI costs squeeze holiday margins.