The AI Tollbooth Finally Pays for the Papers
Google’s News AI pilot has more than 400 participating publications. Payment is real; the traffic promise still needs a scoreboard.
Read the pick: The AI Tollbooth Finally Pays for the PapersX OPPORTUNITY RADAR / THE BRIEF
An EU court adviser says some real-time chatbot reuse may infringe press rights. He recommends declining the training questions, while offering an alternative text-and-data-mining analysis. The Court has not ruled.

The first serious legal line in Europe’s Gemini press-copyright case is narrower than either side’s preferred slogan.[1][2]
On November 3, Advocate General Maciej Szpunar issued an opinion in Case C-250/25, Like Company v Google Ireland. He said an AI chatbot’s real-time retrieval and reuse of an online press publication to answer a user may, in some circumstances, infringe the press publisher’s rights. The point turns on protected expression, not on ownership of facts.[1][3]
His analysis considers unlawful access, including bypassing technical restrictions, and whether an answer recognizably reuses the publication’s expression extensively enough to substitute for reading it. Mere facts are not protected; systematically recognizable reuse also matters to the separate making-available right. Provider liability depends on those uses being inherent to its business model or inadequate effective preventive measures.[1]
Szpunar recommended treating the training questions as inadmissible: the record did not establish use of Like Company’s articles for training, and training is not online use of press publications. His alternative discussion nevertheless supports applying the text-and-data-mining exception. Real-time answer generation and earlier model training remain distinct legal acts.[1]
Article 4 requires lawful access and permits rights holders to reserve this use appropriately, including through machine-readable means for online content. It is a conditional exception, not a blanket free pass for anything a crawler can reach.[4]
This is an Advocate General’s opinion, not a judgment. The Court’s judges are not bound by it. Case C-250/25 remains pending, and the final ruling may follow, narrow or reject the proposed analysis.[1][3]
The underlying dispute was referred by a Hungarian court after Like Company challenged Gemini outputs that allegedly reproduced or summarized material from its news site. The Court’s official case record asks whether chatbot outputs can amount to communication or reproduction of press publications, whether training is reproduction, whether the text-and-data-mining exception applies, and when responsibility lies with the service provider.[2]
Our read: the opinion rejects the lazy fiction that every AI use is one legal blob. Training, retrieval and output can trigger different rules. That makes the compliance work harder, which is usually when slogans begin looking for an emergency exit.
Correction ·
Added the recommended inadmissibility of the training questions, the alternative nature of that analysis and the conditions on output reuse, provider responsibility and the text-and-data-mining exception. Replaced a general Court page citation with the precise press release.
Checked Nov 3, 2026. Numbered references point to the evidence behind factual passages. Commentary is our own.