Short answer: On 3 November 2026 Advocate General Maciej Szpunar gave his opinion in Case C-250/25, Like Company v Google Ireland, the first generative-AI case to reach the EU Court of Justice. He advised that a chatbot that retrieves a press article and reuses it to answer a user may infringe the press publisher’s right if it got the article unlawfully, for example by getting around a paywall, or if the answer recognisably reuses the article’s form of expression so that the reader need not open the original. On training, he said the text-and-data-mining exception generally applies, but treated the training questions as inadmissible. The opinion is not binding, and the Court has not yet ruled. Publishers and writers can keep dated copies of what they published and their access settings. A TimeBind dates the files you hold. It does not decide whether anyone infringed.
Last checked 13 November 2026. General information for publishers, journalists and writers, not legal advice. This page summarises the Court of Justice press release and a law firm’s commentary. We have not read every paragraph of the opinion, which is long, and the opinion is not a judgment.
Table of Contents
What the opinion says
Like Company, a Hungarian operator of web portals, says Google used its articles to train the Gemini models and that Gemini reproduced and made them available to the public between 13 June 2023 and 31 May 2024. The Budapest Regional Court sent four questions to the Court of Justice, two about training and two about the chatbot’s answers.
According to the Court’s press release, the Advocate General advised that the two training questions be declared inadmissible because there was no evidence Like Company’s articles were used to train Gemini and because training is not “online use” of a press publication. For completeness he added that using protected works to train a model is, as a rule, freely permitted under the text-and-data-mining exception.
On answers, he said the press publisher’s right protects the specific form of expression, not the facts reported. Reproduction may occur if the chatbot accessed the publication unlawfully, in particular by circumventing effective technical measures such as a paywall, or if the answer reuses the form of expression so recognisably that the user can dispense with the original. A law firm’s summary adds that, in his provisional view, Like Company’s rights were not infringed on the facts, which is for the Hungarian court to decide.
What it does not settle
An Advocate General’s opinion is advice to the judges. The Court of Justice has not given its judgment, and it may follow the opinion or not. The opinion concerns the EU press publishers’ right under Article 15 of the 2019 Digital Single Market Directive, not copyright in every kind of work. Law firm commentary notes that other referrals, including two other Hungarian cases against Google, are also pending. The opinion does not say Gemini, or any other tool, infringed anyone’s rights in practice.
It also does not say what a writer outside the EU can do, or what applies in other cases such as image or music training.
Who is responsible for a chatbot’s output
The law firm’s note says the Advocate General put primary responsibility on the user where the user directs the chatbot to do something copyright-relevant, but also on the provider where it deliberately intervenes, for example where such acts are inherent to its business model or where it has not put in place appropriate measures a reasonably diligent operator would be expected to take. That is the provisional view of one adviser and the language is for lawyers to apply.
Why people are searching for this
Measured on 13 November 2026. A Reuters post on 3 November had 46,655 impressions, 48 likes and 17 reposts. The EU Court of Justice’s own post on the opinion had 6,447 impressions and 37 likes. Law firms, trade outlets and news sites published summaries over the following week, and X’s trending summary on 3 November described the opinion as a possible shift for how AI handles news in Europe. Counts keep moving.
Checklist for publishers and writers
- Keep a dated copy of each piece as published. Save the final text, headline, images and the publication URL and date.
- Record your access controls. Note paywall, robots.txt and any machine-readable reservation of rights in force on the date, with a screenshot or file copy.
- Save logs that show automated access. If your server logs show crawlers or bots, keep the relevant extracts with the date range.
- Keep side-by-side examples with context. If you think an answer reuses your article, save the question asked, the full answer and the date, plus your article text.
- Separate facts from expression. Note which parts are your original wording and structure and which are facts anyone may report.
- Keep your licences and syndication terms. Record who may reuse what, on what terms, as they stood on the date.
- Keep your contracts and credits. If you are a freelance writer, record who owns the piece and where it appeared.
- Fingerprint the set. Date the article files, access-setting captures and examples together before anything changes.
What a TimeBind does and does not show
A TimeBind is a dated, checkable record that an exact file existed no later than a recorded time and still matches. The SHA-256 fingerprint is made in your browser and the file is never uploaded. The record is anchored to Bitcoin (a Merkle root in an OP_RETURN). Your first Pooled TimeBind is free; after that, TimeBinds start from US$5, and an ASAP TimeBind is US$15. See how it works.
For a publisher or writer, it can date a published piece, a robots.txt or paywall capture and a saved chatbot answer. It does not show that something was copied, who copied it, that an answer was reproduced from your article, that you own the rights, or that a tool used or did not use your work for training. It is one dated record to put next to legal advice.
Limits
This page summarises the Court’s press release and a law firm’s commentary and has not been checked against every paragraph of the opinion. The legal position may change with the judgment, and the law differs outside the EU. It is not legal advice.
Tools that can help
Use the bulk SHA-256 manifest to fingerprint a folder of files, and the process promise tool to write a plain note about how something was made. For related reading see the New York Times v Brave suit over AI answers and paywalls, what authors should keep after the Anthropic settlement and the shared robots.txt AI training opt-out.
Frequently Asked Questions
-
What did the Advocate General say about Gemini on 3 November 2026?
That when a chatbot retrieves a press article and reuses it to answer a user, the publisher’s right may be infringed in certain circumstances, such as unlawful access around a paywall or an answer that recognisably reuses the article’s expression so the user can skip the original.
-
Did the opinion say AI training is allowed?
He said, for completeness, that using protected works to train a model is as a rule freely permitted under the EU text-and-data-mining exception, while advising the Court to treat the training questions as inadmissible. That is a non-binding view.
-
Is the opinion a ruling?
No. It is advice to the Court of Justice, which has not yet ruled. The Hungarian court decides the facts.
-
Does a TimeBind show a chatbot copied my article?
No. It shows that exact files existed no later than a recorded time and still match. It does not show copying, ownership or training use.
-
What should a publisher keep?
Dated copies of what was published, records of paywall and robots.txt settings, relevant server log extracts, and any saved chatbot answers with the question and date.
Sources checked
- Court of Justice of the EU, Press release No 155/26, “Advocate General Szpunar: the retrieval and reuse, in real time, of online press publications by AI chatbots…” 3 November 2026, https://curia.europa.eu/site/upload/docs/application/pdf/2026-11/cp260155en.pdf (read 13 November 2026)
- Opinion of Advocate General Szpunar in Case C-250/25, Like Company v Google Ireland Limited, 3 November 2026 (text via Courthouse News PDF, read 13 November 2026)
- Slaughter and May, “AG Szpunar opines on interplay between AI and press publishers’ rights in Like Company v Google,” 10 November 2026, https://thelens.slaughterandmay.com/post/102o8v0/ag-szpunar-opines-on-interplay-between-ai-and-press-publishers-rights-in-like-co (read 13 November 2026)
- X posts by Reuters and the EU Court of Justice (@EUCourtPress), 3 November 2026; impression counts as shown on 13 November 2026
Date your published pieces and access settings
Your first Pooled TimeBind is free, with no credit card required. After that, TimeBinds start from US$5, and an ASAP TimeBind is US$15. Click here to start, or see plans and pricing.
