The Pulse
Hungarian Court Sends Third Gemini Copyright Case to EU Judges
A Hungarian court has referred a third Google Gemini copyright dispute to the EU’s Court of Justice. Judges will examine how copyright rules apply to AI training, chatbot responses and publishers’ technical opt-outs.
AI.info Team · Written with AI ·
A third Hungarian copyright referral over Google’s Gemini chatbot puts two separate uses of publishers’ work before the European Union’s top court: training an AI model and generating answers for users. In a report published October 12, MLex said judges are seeking guidance on whether those activities fall under the same rules, how copyright exceptions apply and whether publishers must use technical tools to opt out.
Three referrals, questions beyond training
The referral adds another case to the Court of Justice of the European Union’s developing review of copyright claims involving Gemini. It does not establish that Google infringed copyright. The EU court interprets EU law; the Hungarian court will have to apply that interpretation to the dispute before it.
The questions described by MLex go beyond whether material may be used to train a model. They also concern the legal treatment of Gemini’s user-facing responses and the reach of exceptions that can permit text and data mining. For publishers, the opt-out issue is practical as well as legal: the court may have to consider whether rights holders must signal their objection through technical means.
Training and answers raise distinct issues
EU copyright rules include a text-and-data-mining exception for uses of protected material, with a mechanism for rights holders to reserve their rights. The European Commission’s Intellectual Property Helpdesk describes that opt-out condition in its account of the earlier Like Company v Google Ireland case.
That earlier case, C-250/25, also came from Hungary. Like Company, a Hungarian press publisher, alleged that Gemini displayed substantial portions of its publications in response to prompts and that its material had been used to train the model. The Court of Justice’s summary of the referral sets out questions about whether chatbot output can amount to communication to the public, whether training involves reproduction and whether text-and-data-mining rules apply.
The new referral raises a related set of legal questions, but the issues should not be collapsed into one. A model’s training process and the text returned in response to a prompt are different activities; the court’s interpretation could affect each differently. The opt-out question adds another layer: a formal right to reserve content matters to publishers only if they can exercise it in a workable way.
Publishers await a legal test, not a ruling on liability
The Court of Justice has not ruled in the new dispute, and the referral itself is not a finding against Google. Its task is to clarify the EU-law questions sent by the Hungarian judges. The national court will then decide the underlying claims in light of that guidance.
For publishers, the stakes reach beyond whether a specific Gemini response reproduces protected material. The court’s answers may clarify how rights reservations interact with AI training and whether technical opt-out signals provide a legally meaningful route to object. For developers, the questions could shape how they assess protected content across both model development and the generation of responses.
The referral leaves the central boundaries unsettled: how EU copyright rules distinguish training from chatbot output, how far text-and-data-mining exceptions extend, and what counts as an effective publisher opt-out. Those are the questions now before the EU judges.