🔍 Read the full analysis: Publishers’ Writing And AI Chatbots: The Copyright Issue In Europe on ThorstenMeyerAI.com
Get business pricing on monitors, keyboards and dev gear
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
On 3 November 2026, Advocate General Maciej Szpunar advised the EU Court of Justice that chatbots can infringe press publishers’ rights in some circumstances, including by bypassing access barriers or reproducing recognisable expression. He said those conditions do not appear to be met in Like Company’s case against Google; the Court has not yet ruled.
Advocate General Maciej Szpunar advised the European Union’s Court of Justice on 3 November 2026 that chatbots which retrieve and reuse online news articles may infringe press publishers’ rights in some circumstances. In a case involving Google’s Gemini chatbot and Hungarian publisher Like Company, he said infringement could arise if a system gets around access restrictions or reproduces an article’s distinctive expression closely enough that users can forgo the original; he said those conditions do not appear to apply to the example before him.
The opinion concerns Case C‑250/25, Like Company v Google Ireland, referred to the Court of Justice by a Hungarian court. Like Company, which operates several Hungarian news portals, alleges that Google used its articles to train Gemini and reproduced them in chatbot answers between 13 June 2023 and 31 May 2024. Those are the publisher’s claims, not findings by the EU court.
The dispute was prompted by a Gemini answer summarising a July 2023 Hungarian news article about a singer’s reported plan to bring dolphins to Lake Balaton. Szpunar’s proposed test focuses on the way the chatbot obtains and presents the material: access obtained by circumventing a paywall or anti-scraping measure could infringe, as could an answer that preserves recognisable elements of the article’s expression. Repeating the underlying facts alone would not meet that second test.
Szpunar said the available information suggests the example does not meet either condition: the article was not protected by an access barrier, the sample answer did not appear to reproduce recognisable wording, and Google appeared to have safeguards. His opinion is advisory, not binding; the Court of Justice will issue the judgment, and the Hungarian court will ultimately apply EU law to the dispute.
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 Line Between Facts and Expression
The opinion could shape how AI search and chatbot services handle journalism across the EU. A rule against bypassing technical or contractual access restrictions would put pressure on providers to respect barriers that publishers use to control access to their sites. A rule against answers that substitute for the original article could also affect how much detail chatbots quote or closely paraphrase.
At the same time, Szpunar’s reasoning does not give publishers ownership of news facts. If the Court adopts the proposed approach, a chatbot could still report the same events in its own wording, provided it has obtained material lawfully and does not reproduce the article’s protected expression in a way that replaces the source. That distinction matters to readers, publishers and AI companies: it could preserve factual summaries while supporting links, licensing discussions and limits on near-copy answers.
The practical consequences remain unsettled until the judges rule. The opinion may inform negotiations and product choices, but it does not itself establish a final EU-wide legal test.
Popular right now
Dyson V8 Cyclone Cordless Vacuum
Bose QuietComfort Headphones (2nd Gen), Wireless Over-Ear Noise-Cancelling Headphones, Hazelnut Taupe
Sony BRAVIA XR8B 65-Inch 4K OLED Google TV (K-65XR8B)
As an affiliate, we earn on qualifying purchases.
How the Hungarian Case Reached Luxembourg
EU law has recognised a related right for press publishers since the 2019 Copyright in the Digital Single Market Directive. Article 15 gives publishers rights over certain online uses of their press publications. The opinion recalls that the measure was adopted with services such as news aggregators in mind; the current dispute asks whether it also reaches chatbot answers that draw on articles.
The Hungarian court sent the EU court four questions: two about whether model training engages the publishers’ right and two about chatbot responses. The hearing took place on 10 March 2026, after written observations from nine EU member states and the European Commission. The EU court described the case as its first generative-AI case.
On model training, Szpunar’s main conclusion was procedural: he advised that the two questions should be treated as inadmissible because the record did not establish that Like Company’s articles were used to train Gemini. He pointed to the article’s publication date and Gemini 1.0’s later release, and said the example appeared instead to involve live retrieval from the web, a method known as retrieval-augmented generation. Separately, and for completeness, he said training is in principle covered by the EU’s text-and-data-mining exception, which allows rights holders to opt out in a machine-readable form. That discussion was not a ruling on whether these articles were used for training.
The Court Has Yet to Rule
The Advocate General’s opinion is not a final judgment, and the Court of Justice has not announced a date for its ruling. The judges may follow, adapt or reject his proposed approach. The Hungarian court must still decide the underlying dispute using the EU court’s eventual interpretation.
It also remains unresolved whether the articles were included in any training data: Szpunar said the record did not establish that they were. His comments on the text-and-data-mining exception were made for completeness after he recommended that the training questions be dismissed as inadmissible. They should not be read as a finding that Google trained Gemini on Like Company’s material or as a final decision on every training-related copyright question.
The opinion’s proposed standards also leave practical questions for later cases, including how courts would identify a recognisable reproduction and determine when an answer lets users avoid the original. The source material reports no settled, binding answer to those questions.
Judgment and Hungarian Proceedings
The next legal milestone is the Court of Justice judgment in Case C‑250/25. No date has been set. The Court’s 2025 statistics put the average duration of preliminary-reference cases at 16.9 months, but that figure is a historical average, not a timetable for this case.
After the EU court answers the questions of EU law, the Hungarian court will consider Like Company’s claims against Google. Until then, the opinion gives publishers and chatbot providers a proposed framework rather than a binding outcome: pay attention to access restrictions, avoid answers that reproduce protected expression as a substitute for the article, and distinguish live retrieval from claims about model training.
Key Questions
Did the EU Court of Justice rule that Google infringed copyright?
No. Szpunar issued a non-binding opinion on 3 November 2026. The Court of Justice has not yet delivered its judgment, and the Hungarian court will decide the underlying case.
When could a chatbot answer infringe publishers’ rights under the opinion?
Szpunar said infringement may arise if a chatbot obtains an article by circumventing an access restriction, or if its answer reproduces the article’s recognisable expression so closely that users can dispense with the original. Reporting the same facts alone would not meet that proposed test.
Did the opinion establish that Gemini was trained on Like Company’s articles?
No. Szpunar said the case record did not establish that the articles were used for training and advised that the training questions be dismissed as inadmissible. His additional discussion of text and data mining was made for completeness, not as a finding about Google’s training data.
What happens next in the case?
The Court of Justice must issue its judgment on the Hungarian court’s questions. It has not set a date. The Hungarian court will then continue the dispute and apply the EU court’s interpretation.
Source: ThorstenMeyerAI.com
