🔍 Read the full analysis: An EU Court Adviser Says AI Chatbots Can Infringe Publishers’ Rights. Facts Stay Free, The Writing Does Not. What It Means For Companies, Publishers And Users on ThorstenMeyerAI.com
Get the latest gadgets 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 Nov. 3, 2026, Advocate General Maciej Szpunar advised the EU Court of Justice that chatbots retrieving and reusing news articles may infringe publishers’ rights if they bypass access barriers or reproduce an article’s distinctive expression. He said the test does not appear to be met in Like Company’s case against Google, and his opinion is not binding. The Court has yet to rule.
Advocate General Maciej Szpunar advised the European Union’s top court on Nov. 3 that AI chatbots that retrieve online news articles and reuse them in answers may, in certain circumstances, infringe press publishers’ rights. In the dispute between Hungarian publisher Like Company and Google, however, he said the conditions for infringement do not appear to be met; the opinion is not a binding judgment.
The case, C‑250/25, Like Company v Google Ireland, concerns a Gemini response summarising a Hungarian article about a singer’s reported proposal to bring dolphins to Lake Balaton. Like Company, which operates several Hungarian news portals, sued Google, alleging that its articles were used to train Gemini and were reproduced or made available through the chatbot from June 13, 2023, to May 31, 2024. Those are the publisher’s allegations, not findings by the Court of Justice.
Szpunar’s proposed test focuses on how the chatbot obtained and reused an article. An answer could infringe if the service gained access unlawfully—for example, by getting around a paywall or anti-scraping measure—or if it reproduced the article’s expression so recognisably that a user could avoid reading the original publication. Summarising the same underlying facts, without taking the article’s protected expression, would not by itself meet that test.
The adviser said the available facts do not appear to show either condition in this dispute: he cited no apparent access barrier, no recognisable wording in the example answer and safeguards that Google appears to have in place. The Hungarian court must decide the case after the EU court answers the legal questions sent to it.
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 Narrow Test for Chatbot Answers
The opinion offers publishers and AI services a framework for disputes over chatbots that search the web and generate answers from current material. It suggests that access controls and the wording of an answer may matter more than the mere fact that a chatbot reports information also found in a news article. That distinction could shape how companies design retrieval systems and how publishers document and enforce restrictions on access.
For publishers, the proposed approach offers a possible legal route against bypassing technical barriers or answers that function as substitutes for articles. It does not, however, give publishers ownership of the facts reported in their stories. Readers may still receive summaries of news, provided those summaries do not unlawfully obtain the material or reproduce its protected expression in the way described by the adviser.
The opinion also leaves the practical boundary open to later court decisions. What counts as expression recognisable enough to replace an original article, and how safeguards should work across different systems, will depend on the evidence and the Court’s eventual ruling. For now, Szpunar’s analysis is guidance to the judges, not a final rule.
Products Worth Considering
LG UltraGear 32GX850A-B 32-inch 4K UHD Glossy OLED Gaming Monitor
EcoFlow Delta 3 Portable Power Station – 1024 Wh, 1800 W
Samsung Galaxy Tab A11+ (6 GB RAM, 128 GB storage, expandable) – Gray
As an affiliate, we earn on qualifying purchases.
How the Gemini Dispute Reached Luxembourg
Like Company brought its claim in Hungary under the press publishers’ right in Article 15(1) of the EU’s 2019 Copyright in the Digital Single Market Directive. That right allows publishers to control certain online uses of their press publications. The directive was adopted with online services such as news aggregators in mind, and the case asks how that right applies to chatbot answers.
The Budapest regional court referred four questions to the EU Court of Justice: two about AI training and two about chatbot responses. The hearing took place on March 10, 2026, after written observations from nine EU member states and the European Commission, according to the opinion. The example at the centre of the dispute was a Hungarian article published on July 21, 2023, and a Gemini summary generated after a user asked about the site and article.
Szpunar treated the training questions separately from the chatbot-answer questions. He advised the Court to find the training questions inadmissible because the record did not support that the article had been used to train Gemini. He noted that the timing made that claim implausible in this instance and said the chatbot appeared to have retrieved the article live, a method known as retrieval-augmented generation. His additional discussion of training was expressly offered for completeness, not as the basis for deciding the case.
The Court Has Yet to Decide
The Court of Justice has not issued its judgment, and no date for a ruling has been announced. Advocate General opinions are not binding; the judges may follow the advice, adopt part of it or reach a different interpretation. The Hungarian court will then apply the EU court’s legal guidance to the evidence in the case.
Several issues remain unresolved, including whether the record supports Like Company’s allegations about Google’s conduct across the period it cited. The opinion’s conclusion about the example response is provisional, not a final finding that Google did or did not infringe. The scope of any safeguards, and what wording would be sufficiently recognisable to replace access to an original article, also remain matters for assessment in future cases.
The training discussion should not be read as a ruling that all AI training is lawful. Szpunar advised that the questions were inadmissible because the case record did not establish use of the articles for training. His separate view that the EU text-and-data-mining exception generally covers training was additional analysis; rights holders may opt out in a machine-readable form, under the directive’s framework.
Judges Consider the Opinion
The Court of Justice will now consider the questions referred by the Hungarian court and issue its judgment in Case C‑250/25. No judgment date has been set. The Court’s 2025 statistics put the average duration of preliminary-reference proceedings at 16.9 months, but that figure does not establish when this specific case will be decided.
After the EU court rules, the Hungarian court is expected to resolve the underlying dispute using that interpretation and the case evidence. Until then, the adviser’s proposed distinction—between reporting the same facts and unlawfully accessing or recognisably reproducing an article—remains guidance rather than a binding outcome.
Key Questions
Has the EU Court ruled that Gemini infringed publishers’ rights?
No. Szpunar issued a nonbinding opinion, not a judgment. The Court of Justice has yet to rule, and the Hungarian court must decide the dispute.
When could a chatbot answer infringe under the adviser’s test?
It may infringe if the chatbot obtains an article by bypassing an access barrier, or if its answer reproduces the article’s expression so recognisably that users can avoid accessing the original. Repeating facts alone is not enough under the opinion’s proposed test.
Did the opinion find that Google trained Gemini on Like Company’s articles?
No. Szpunar advised that the training questions were inadmissible because the record did not establish that the articles were used for training. He said the example appeared to involve live retrieval of the article, rather than use of it as training data.
Does the opinion establish that AI training is lawful?
No. The adviser’s main procedural conclusion was that the training questions should not be decided on the evidence presented. His discussion of the EU text-and-data-mining exception was supplementary and does not amount to a binding ruling on training practices.
Source: ThorstenMeyerAI.com
