Case C-829/26, Mobilguru – training, text and data mining, and the robots.txt file
The press publishers’ right covers “online use” by “information society service providers”. The Budapest court asks whether training an AI model is such a use, which is the question Advocate General Szpunar answered in the negative two weeks ago.
Facts
Mobilguru Kft. is suing Google Ireland Limited before the Fővárosi Törvényszék (Budapest-Capital Regional Court) in a dispute about the press publishers’ right in Article 15 of Directive (EU) 2019/790. The notice gives no further facts. The referring court asks eleven questions, several with sub-questions. They concern four matters: whether the “service” that Article 15 presupposes includes the training of the model or only its answers to users; what the press publishers’ right covers at each stage; how the text and data mining exception in Article 4 interacts with that right; and how a publisher may validly reserve its rights under Article 4(3), including by a robots.txt file. It is the third Hungarian reference against Google on chatbots and press publications, after Like Company (C‑250/25) and Manchester United Fanatics Klub (C‑806/26).
Questions Referred
The principal questions read as follows; questions 5 to 10 and the remaining sub-questions, on the relationship between the text and data mining exception and the press publishers’ right, are set out in full in the Official Journal notice.
(1) Must ‘service’, within the meaning of Article 2(a) of the Directive on electronic commerce and Article 1(b) of Directive 2015/1535, be understood to mean solely the mechanisms by which AI models respond to users’ prompts (strict interpretation), or – given that it is the same provider that creates the AI models and provides the service that makes use of them – must that concept also be considered to include the training of the AI [model], that is to say, the activities carried out during its creation (broad interpretation)?
(2) Is it correct to interpret the first subparagraph of Article 15(1) of Directive 2019/790 as meaning that the right of publishers of press products to authorise and prohibit, in accordance with Article 2 and Article 3(2) of Directive 2001/29/EC, the reproduction and the making available of those products to the public ‘on-demand’ is limited exclusively to online use by information society service providers?
[…]
(4) If, depending on the answer to the first question referred, operations carried out during the training of the AI model are also considered to be an information society service, that is to say, that the broad interpretation is the correct one, must Article 15(1) of Directive 2019/790 be interpreted as meaning that, in accordance with that article, the right of press publishers to authorise and prohibit the use of their press publications – with the exceptions provided for in Article 15(1) and (3) – also includes acts of reproduction potentially carried out during the training of the AI model?
[…]
(11) By way of example, do the following constitute an appropriate means of expressing a reservation in relation to the authorisation of reproduction for the purposes of text and data mining, within the meaning of Article 4(3) of Directive 2019/790 and of Article 53(1)(c) of Regulation 2024/1689: — a reservation expressed in the robots.text file of the internet page of an online press product, prohibiting access to the internet page in a machine-readable manner; — making public, without recourse to the robots.txt file, but in any other part of the press product itself – whether in the legal information, on the home page of the press product or below the title – that use for the purposes of text and data mining or for other purposes is prohibited?
Sources
OJ notice C/2026/5881 (EUR‑Lex) · Case file on CURIA · Opinion of AG Szpunar in Like Company, C‑250/25 (ECLI:EU:C:2026:919) · Directive (EU) 2019/790 · Regulation (EU) 2024/1689
Comment
The first question has, in effect, already been answered by the Advocate General, if not yet by the Court. In his Opinion of 3 November 2026 in Like Company, Advocate General Szpunar proposed that training is not an “online use” of a press publication within Article 15. It “occurs as part of the development of a piece of AI technology on dedicated computer infrastructure, without any information society service being provided to a user via the internet”. The fact that the same provider later offers a chatbot does not, in his view, change that, because the link between the “upstream” use and the “downstream” answer is too indirect. The Budapest court frames the point as a choice between a “strict” and a “broad” reading of “service”. If the Court follows the Advocate General, the strict reading wins, and questions 4 and 5 fall away.
That would not end the case. On the strict reading, the second question asks who controls uses that fall outside Article 15: if the press publisher has no right over training, the authors of the articles and other rightholders may still have one. Training would then be governed by ordinary copyright and the text and data mining exception in Article 4, not by the publishers’ right. The press publishers’ right would be confined to what the chatbot does when it answers.
The eleventh question is the one with the widest practical effect. Article 4(3) allows rightholders to exclude text and data mining by reserving their rights “in an appropriate manner, such as machine-readable means in the case of content made publicly available online”. Article 53(1)(c) of the AI Act requires providers of general-purpose AI models to “identify and comply with, including through state-of-the-art technologies”, such reservations. The Budapest court asks whether a robots.txt entry blocking access is enough, and whether a notice in the site’s legal information, on its home page or under the headline will do. In Like Company the Advocate General suggested that a reservation is appropriate where “state-of-the-art robots and associated technologies are capable of ‘understanding’ it with a sufficient degree of reliability at the moment of access”. A notice in the legal small print would struggle to meet that test. Whether the Court adopts it will decide how every publisher in the Union has to opt out.