🔍 Read the full analysis: Can AI Use Facts Without Using A Publisher’s Protected Writing? on ThorstenMeyerAI.com
Get smart everyday buys delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
On November 3, 2026, Advocate General Maciej Szpunar advised the EU Court of Justice that chatbots retrieving and reusing press articles may infringe publishers’ rights in certain circumstances. He said reporting the same facts is not enough: unlawful access or answers that closely reproduce an article’s expression may cross the line. The opinion is non-binding, and the Hungarian court must still decide the dispute.
EU Court of Justice Advocate General Maciej Szpunar said on November 3 that chatbots which retrieve online news articles and reuse them in answers may infringe press publishers’ rights if they access material unlawfully or reproduce its distinctive wording closely enough to substitute for the original. His opinion in Case C‑250/25, Like Company v Google Ireland is advisory, not a final ruling, and says the test does not appear to be met by the example at issue: a Gemini summary of a Hungarian article about dolphins.
The dispute began after Like Company, which operates Hungarian news portals, sued Google in Hungary. The company alleged that Google used its articles to train Gemini and that the chatbot repeatedly reproduced and made them available to users from June 13, 2023, through May 31, 2024. Those are Like Company’s allegations; the Advocate General’s opinion does not establish that the articles were used for training.
The Hungarian court referred four questions to Luxembourg: two concerning AI training and two about chatbot answers. Szpunar advised that the training questions were inadmissible because the record did not show that the articles were part of Gemini’s training data. He reasoned that the example article was published in July 2023, after the model’s development was underway, and that the response likely came from a live web search rather than training. The opinion describes this as retrieval-augmented generation, in which a system retrieves pages and generates an answer from them.
For chatbot answers, the proposed test focuses on access and expression. An infringement may arise if a service gets around a paywall or anti-scraping measure, or if its response repeats the article’s expressive form so recognisably that a user can do without reading the original. Repeating facts alone is not sufficient, according to the opinion. Szpunar said the example did not appear to involve access barriers or recognisable copied wording, and that Google appeared to have safeguards.
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
A Line Between Facts and Copying
The opinion addresses a central practical question for news publishers and AI services: whether a chatbot can use reporting to answer a question without reproducing protected expression or bypassing the publisher’s controls. Szpunar’s approach would leave facts available for discussion while giving publishers a potential basis to challenge certain ways articles are retrieved or closely paraphrased.
For publishers, the view could support arguments for licensing when a service uses access-restricted material or produces answers that act as substitutes for articles. For chatbot operators, it points to the importance of respecting technical access controls and keeping generated responses distinct from source wording. The opinion does not establish a general right to payment for every AI answer or give publishers ownership of the underlying news facts.
Users could be affected if services change how they retrieve articles, summarize them, or link to original reporting. Those effects remain possibilities, not outcomes established by the opinion. The Court of Justice’s eventual judgment and the Hungarian court’s application of the law will determine what the test means in this dispute.
Products Worth Considering
Dell UltraSharp 32 U3226QER 4K Thunderbolt Hub Monitor
Samsung 27-inch Odyssey G5 QHD Gaming Monitor (G51F)
Samsung 34-Inch Odyssey G5 Ultra-Wide Curved Gaming Monitor
As an affiliate, we earn on qualifying purchases.
How the Hungarian Case Reached Luxembourg
Like Company brought its case before a regional court in Hungary, which asked the EU Court of Justice to interpret the press publishers’ right under Article 15 of the 2019 Copyright in the Digital Single Market Directive. The right was introduced in part to let publishers control or license certain online uses of press extracts, including by news aggregation services. The current case asks how that framework applies to generative chatbots.
The Court of Justice’s press release describes this as the first generative AI case to reach the court. Nine EU governments and the European Commission submitted written observations, and the hearing took place on March 10, 2026. An Advocate General’s opinion provides independent legal advice to the court; judges may follow it, but they are not bound by it.
On training, Szpunar’s main conclusion was procedural: the questions should not be answered because the case record did not establish that the articles were used to train Gemini. He added, as a separate discussion, that training may generally fall within the EU text-and-data-mining exception, subject to rights holders’ ability to reserve their rights in a machine-readable way. That discussion was not a ruling on whether Google trained Gemini on Like Company’s material.
The Court Has Not Ruled Yet
Szpunar’s opinion is not binding, and the Court of Justice has not set a date for its judgment. The judges may adopt, modify or reject his reasoning. The Hungarian court will then have to decide the case in light of the EU court’s interpretation and the evidence before it.
It also remains unresolved whether Like Company’s articles appeared in Gemini’s training data. The opinion says the record did not support that claim, rather than making a final factual finding about all of Google’s training practices. Nor does the opinion settle how courts should assess every kind of AI summary, paraphrase, technical barrier or licensing arrangement.
A Judgment Before the Case Ends
The next major step is a judgment from the Court of Justice in Case C‑250/25. The court’s statistics show that preliminary-reference proceedings took an average of 16.9 months in 2025; that figure is historical and does not set a timetable for this case. After the EU court rules, the Hungarian court will resume the dispute and apply the interpretation to the parties’ claims and evidence.
Until then, the opinion offers guidance rather than a final legal outcome. The key points to watch are whether the judges adopt Szpunar’s distinction between facts and protected expression, how they treat access barriers, and whether they address the separate training questions despite his recommendation that they be dismissed.
Key Questions
Has the EU Court of Justice ruled that Gemini infringed copyright?
No. Szpunar issued a non-binding opinion; the court has not yet delivered its judgment. He said the conditions for infringement did not appear to be met in the example described.
Can a chatbot repeat facts from a news article?
According to the opinion, reporting the same facts is not enough by itself to infringe the press publishers’ right. The proposed test focuses on unlawful access or recognisable reuse of the article’s protected expression.
Did the opinion find that Google trained Gemini on Like Company’s articles?
No. Szpunar said the record did not show that the articles were used for training and advised that the training questions were inadmissible. His additional discussion of text-and-data-mining rules was not a finding about Google’s training data.
What happens after the Advocate General’s opinion?
The Court of Justice will issue its own judgment, with no date announced in the supplied material. The Hungarian court will then decide the underlying dispute using the EU court’s interpretation.
Source: ThorstenMeyerAI.com
