legal247

Press publishers' right and AI chatbots: what does the Advocate General say in C-250/25?

In brief

The press publishers' right under Article 15 of the DSM Directive may be infringed when an AI chatbot retrieves a press article in real time and reuses its expression in an answer, according to Advocate General Szpunar in C-250/25 Like Company. Decisive are unlawful access and whether the answer replaces the original. Training is, in his view, as a rule permitted under the TDM exception.

Advocate General Maciej Szpunar delivered his Opinion in Case C-250/25 Like Company v Google Ireland on 3 November 2026. The case is the first concerning generative AI to reach the Court of Justice, and it concerns whether Google Gemini infringes the right that press publishers hold under Article 15 of the DSM Directive when the chatbot retrieves and reuses press articles to answer users. The Advocate General considers that such use may in certain circumstances amount to an infringement, and he sets out the conditions under which that is the case. The Opinion is relevant to press publishers, to providers of AI services and to organisations building their own chatbots with web search or retrieval from external sources.

What is Case C-250/25 Like Company v Google about?

Like Company operates several web portals in Hungary and brought proceedings against Google before the Budapest Regional Court. The company claims that Google used the articles on its portals to train the language models behind Gemini, and that the chatbot repeatedly reproduced the articles and made them available to the public between 13 June 2023 and 31 May 2024. Gemini uses retrieval-augmented generation, meaning that the chatbot collects information from the web in real time to improve its answers. The Hungarian court referred four questions for a preliminary ruling. Two concern the training of language models, and two concern the answers the chatbot gives to users.

According to the Court’s press release, the Advocate General also raises, without resolving it, the question whether Like Company’s portals are press publications within the meaning of the Directive at all. That question affects the outcome, but it is for the national court to decide.

What is the press publishers’ right under the DSM Directive?

The press publishers’ right is a related right giving publishers of press publications exclusive rights of reproduction and making available when information society service providers use their publications online. Under DSM Directive Art. 15(1), publishers are granted the rights in InfoSoc Directive Arts. 2 and 3(2) for such use. The right does not apply to private or non-commercial uses by individual users, to hyperlinking, or to individual words or very short extracts. Under DSM Directive Art. 15(4), the right expires two years after the publication is published, calculated from 1 January of the following year. Authors of works incorporated in the publication are to receive an appropriate share of the revenues under DSM Directive Art. 15(5).

The right was designed with news aggregators and search engines in mind, and it has been unclear whether it also covers use in AI services. The Opinion in C-250/25 is the first indication of how the Court of Justice may answer that question.

Do AI chatbots infringe the press publishers’ right by real-time retrieval?

The Advocate General considers that the press publishers’ right may be infringed when a chatbot driven by a language model retrieves a publication from the publisher’s website, reuses it to generate an answer and provides that answer to the user. He stresses that the right does not protect the facts reported in the publication, only its specific form of expression. Its aim is to protect the investment that goes into the publication, not to give publishers a monopoly over the news. Whether the reproduction right is infringed depends, in his view, on two factors.

Factor Points against infringement Points towards infringement
Access to the publication Lawful access to openly available content Unlawful access, in particular by circumventing a paywall or technical measures against scraping
Content of the answer The answer merely reports facts from the publication The answer reuses the form of expression recognisably, so that the user need not read the original
Pattern across answers Isolated similarities The form of expression is systematically recognisable in answers to related queries

The right of making available to the public is, according to the Opinion, as a rule infringed where the form of expression of the publication is systematically recognisable in the chatbot’s answers to related queries. The criterion that the answer makes reading the original unnecessary ties the assessment to the economic harm to the publisher. In our assessment, a short summary with a reference to the source will normally fall outside the right, whereas an answer reproducing large parts of an article behind a paywall will lie close to its core.

When is the provider of the chatbot liable?

According to the Opinion, the provider of the chatbot is liable for such uses where they are inherent to its business model, or where it has failed to put in place appropriate measures to counter them credibly and effectively. The standard resembles the one the Court of Justice has developed for platforms that host content, and it places the duty of prevention on the provider of the service. For providers, technical barriers against verbatim reproduction, respect for paywalls and reservations, and procedures for handling complaints from publishers may therefore determine liability.

The Opinion does not expressly address the liability of organisations that build their own solutions on a provider’s model. Where an organisation itself controls the sources from which the chatbot retrieves content, and itself offers the service to the public, the better view in our assessment is that the organisation may be regarded as the user of the publications. This applies in particular to customer service and knowledge assistants that retrieve content from trade press or news services.

Is AI training on press articles permitted under the text and data mining exception?

The Advocate General proposes that the questions on training be declared inadmissible. There is no evidence that Like Company’s articles were used to train Gemini, and training a language model is, in his view, not an online use of press publications, so that it falls outside the press publishers’ right. For the sake of completeness, he adds that the use of protected works and other subject matter to train a language model is, as a rule, freely permitted under the text and data mining exception in DSM Directive Art. 4.

The remark on training is subsidiary and has here been checked only against the press release, so the full Opinion should be read before weight is attached to it. The exception in Article 4 applies only to lawfully accessible works, and it does not apply where the rightholder has reserved its rights in an appropriate manner, for example by machine-readable means. The Opinion therefore does not diminish the significance of reservations. Providers of general-purpose AI models must also have a policy to comply with EU copyright law, including reservations, under AI Act Art. 53(1)(c). The rules on training are discussed further in the article on copyright and AI, and the proposed Norwegian exception and the value of providers’ indemnities are covered in the article on text and data mining.

What does C-250/25 mean for Norwegian organisations?

Norway has not yet implemented the DSM Directive. In Prop. 41 LS (2025–2026), the Government proposed implementing the Directive, including rules on the publisher’s right in press publications and an exception for text and data mining. According to the Storting’s case page, the bill is still before the Standing Committee on Family and Cultural Affairs. Once the rules are adopted, they are to be interpreted in line with the case law of the Court of Justice, and a judgment in C-250/25 will then have direct significance for Norwegian law.

For Norwegian press publishers, the case is relevant already, because AI services are offered in the EU Member States where publishers may rely on the right. The case is also linked to the dispute over Meta’s new terms, in which the Oslo District Court emphasised that the terms gave Meta an extensive right to text and data mining. For organisations using AI services, the main consequence is that the risk of infringement lies in what the chatbot delivers to the user, and not only in how the model was trained.

What should organisations do following the Opinion in C-250/25?

Organisations should assess their AI solutions and contracts against the conditions set out by the Advocate General, even though judgment has not yet been given. Further articles on AI and contracts are available on the artificial intelligence topic page.

  1. Map the external sources from which the organisation’s chatbots and knowledge assistants retrieve content in real time, and whether these include press publications.
  2. Restrict answers technically so that they report facts and the source, and not longer verbatim extracts from articles.
  3. Ensure that the solution respects paywalls, robots.txt and other reservations, and that it does not use credentials it is not entitled to use.
  4. Require the AI provider by contract to have measures against verbatim reproduction and to indemnify the organisation against claims from rightholders.
  5. Enter into a licence with the relevant publishers where the service depends on reproducing articles, for example in a media monitoring service.
  6. For press publishers, adopt a machine-readable reservation against text and data mining and document access restrictions, so that unlawful access can be demonstrated if the content is later reproduced.

Questions and answers

Is the Advocate General's Opinion in C-250/25 binding?

No. The Advocate General proposes a solution to the Court of Justice, which is not bound by it. The judges have begun their deliberations, and judgment will be given at a later date. In practice the Court often follows the Advocate General's outcome, but not always the reasoning.

Does the press publishers' right apply in Norway?

Not yet. The Norwegian Government has proposed implementing the DSM Directive, including a right for publishers of press publications, in Prop. 41 LS (2025–2026). The bill is before the Storting. Norwegian publishers may nevertheless rely on the right in the EU Member States where their articles are used.

Does the press publishers' right protect the news itself?

No. The right protects the specific form of expression of the press publication and not the facts it reports. A chatbot answer that reports the facts in its own words therefore does not infringe the right. Hyperlinks, individual words and very short extracts are also excluded under Article 15(1) of the DSM Directive.

Next legal review: 1 December 2026