🔍 Read the full analysis: Facts Stay Free, Writing Does Not: What That Means For AI Chatbots on ThorstenMeyerAI.com
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
On Nov. 3, 2026, Advocate General Maciej Szpunar advised the EU Court of Justice that chatbots which retrieve and reuse news articles may infringe publishers’ rights in some circumstances. He said reporting the same facts is not enough; unlawful access or reuse of recognizable expression may cross the line. His opinion is nonbinding, and he said the apparent facts in the Gemini case do not meet that test.
Advocate General Maciej Szpunar advised the European Union’s top court on Nov. 3 that an AI chatbot may infringe press publishers’ rights when it retrieves and reuses news articles, but only in specific circumstances. In Case C‑250/25, involving Google’s Gemini and a Hungarian publisher, he said a summary can cross the line if it follows unlawful access to an article or reproduces its recognizable expression closely enough to replace the original; the opinion is not a final ruling.
The dispute began with a July 2023 Hungarian article about a popular singer’s reported plan to bring dolphins to Lake Balaton. Like Company, which operates several Hungarian news portals, sued Google, arguing that Gemini used its articles in training and repeatedly reproduced them in chatbot answers between June 2023 and May 2024. According to the Court of Justice’s press release, a user asked Gemini to summarize the dolphin article, and the chatbot produced a summary in Hungarian.
Szpunar drew a distinction between facts and expression: a chatbot may report the same facts without infringing the publishers’ right, but it may infringe if the article was obtained by bypassing a paywall or anti-scraping measure, or if the response preserves distinctive wording or structure so recognizably that a user can forgo the original. The opinion’s test is therefore not a general ban on AI summaries or on reporting news facts.
On the example before him, Szpunar said the infringement conditions do not appear to be met: the article was not behind an access barrier, the sample answer did not reproduce recognizable wording, and Google appeared to have safeguards. Those are his assessments in a legal opinion, not findings in a binding judgment. The Hungarian court, which referred questions to Luxembourg, will decide the case after the EU court rules.
Infographic 01 · Where the line runs
When a chatbot answer infringes a publisher's right
Advocate General Szpunar's test in Case C‑250/25 for chatbots that retrieve news articles in real time and reuse them in their answers.
Stays free
- Reporting the same facts. Nobody owns the news itself (pts 199–200).
- A short factual summary, presented differently (pt 208).
- An overview built from several sources, restructured, with links back (pts 208, 234).
- Individual words or very short extracts (pt 201).
Can infringe
- Getting around a paywall or anti-scraping block to fetch the article (pt 188).
- An answer that keeps the article’s structure and phrasing, even paraphrased, so readers can skip the original (pts 202, 205, 207).
- Systematically recognisable articles in answers to related questions: making available to the public (pts 213, 240).
The test, step by step
For a chatbot that fetches a news article live (RAG) and answers a user
- Is it an online use by a service in the EU?Yes for RAG answers to EU users, wherever the servers are (pts 165–169).
- Did the bot get in lawfully?No circumvention of effective technical measures (pt 188).
- Is the article’s form of expression recognisable in the answer, to the point of substitution?Side-by-side comparison; facts do not count (pts 196–209).
- Is the provider responsible?If inherent to its business model, or no reasonable measures such as respecting blocks and output filters (pts 221–222).
- Temporary copies during retrieval are then exempt under Art. 5(1) InfoSoc; the TDM exception does not apply to RAG (pts 179–189).
An Advocate General’s opinion, not a judgment. General information, not legal advice.
Sources: Opinion of AG Szpunar, C‑250/25 (provisional text), CURIA press release No 155/26 (3 Nov 2026). Point numbers refer to the provisional text.
Powered by Thorsten Meyer AI
The Boundary Between Summaries and Copies
The opinion offers a possible legal standard for AI search and chatbot products that retrieve current web pages to answer questions. If the court adopts it, services may need to account for how they access publisher sites and how much of an article’s particular expression their answers retain. The opinion does not say that every summary requires a licence, nor does it give publishers control over the underlying facts.
For publishers, the distinction could shape efforts to license or challenge chatbot use of their reporting. For users, the practical effect may be answers that rely more on synthesis across sources and links back to original articles. The precise effect will depend on the court’s eventual judgment and how national courts apply its interpretation to specific products and responses.
Products Worth Considering
EcoFlow Delta 3 Portable Power Station – 1024 Wh, 1800 W
Bose TV Speaker Compact Soundbar with Bluetooth and HDMI-ARC, Black
Anker SOLIX C300 Portable Power Station, 288Wh
As an affiliate, we earn on qualifying purchases.
A Dispute Over Gemini’s Live Search
Like Company brought its claim in a Budapest regional court under the EU press publishers’ right, established in the 2019 Copyright in the Digital Single Market Directive. That right was adopted in part to let publishers control and license certain uses of press extracts by online services, including news aggregators. The question referred to the EU court is how that right applies when a chatbot retrieves an article and uses it to answer a user.
The case also raised whether Google used Like Company’s articles to train Gemini. Szpunar advised the court not to answer those training questions because, in his view, the referral did not establish that the articles were in the training data. He considered it more plausible that Gemini had retrieved the article live through retrieval-augmented generation, in which a system searches for relevant pages and builds an answer using them. The hearing took place on March 10, 2026, and nine EU governments and the European Commission submitted written observations, according to the opinion.
Although he treated the training questions as inadmissible, Szpunar discussed training separately “for the sake of completeness.” He said it involves copying in a legal sense and expressed the view that the EU’s text-and-data-mining exception generally covers such use, subject to rights holders’ ability to opt out in a machine-readable way. That discussion is not the court’s ruling and does not decide whether Google trained Gemini on these particular articles.
The Final Legal Test Is Pending
The Advocate General’s opinion is advisory and nonbinding; the Court of Justice has not issued its judgment, and no judgment date was given in the source material. The court may follow, adjust or reject his reasoning. The opinion also does not establish that Gemini used Like Company’s articles for training, or settle the facts of every chatbot’s access and response practices.
Whether a particular answer reproduces recognizable expression may depend on the article, the chatbot’s wording, the access method and the safeguards in place. The source material does not specify a final legal threshold for measuring how much similarity is enough. Szpunar’s comments on text-and-data mining are his analysis, not a binding determination of the scope of that exception in this dispute.
Court Judgment and Budapest Proceedings
The Court of Justice will issue its judgment in response to the Hungarian court’s questions. The court’s 2025 statistics recorded an average of 16.9 months for preliminary references, but that figure is a historical average, not a scheduled timetable for this case. No date has been announced.
After the EU court answers the legal questions, the Hungarian court will decide the underlying case. That decision will determine how the proposed interpretation applies to Like Company’s claims against Google. Until then, the opinion provides guidance for the dispute but does not settle liability or create a final judgment in the case.
Key Questions
Did the EU court rule that Gemini infringed the publisher’s rights?
No. Advocate General Szpunar issued a nonbinding opinion on Nov. 3, 2026. The Court of Justice’s judgment is still pending, and the Hungarian court will decide the underlying dispute.
When might a chatbot summary infringe press publishers’ rights?
Under Szpunar’s proposed test, infringement may arise if a chatbot obtains an article by bypassing an access barrier, or if it reproduces recognizable expression so closely that readers can avoid accessing the original. Repeating the same facts alone is not enough under that test.
Did Szpunar find that Google trained Gemini on the articles?
No. He said the training questions should not be answered because the case materials did not establish that the articles were used for training. He considered live retrieval a more plausible explanation for the example summary.
Does the opinion mean AI training is always allowed?
No. Szpunar discussed the EU text-and-data-mining exception as a general matter, but that discussion was not a binding ruling and did not establish that Google trained Gemini on Like Company’s articles. He also referred to rights holders’ ability to opt out in a machine-readable form.
Source: ThorstenMeyerAI.com
