Like Company v. Google

Like Company, a Hungarian press publisher that operates several online news portals, brought proceedings against Google Ireland Limited before the Budapest Környéki Törvényszék (Budapest Regional Court). It alleges that Google’s Gemini chatbot (formerly Bard) reproduced and made available to the public excerpts of its copyright-protected news articles without a license. The alleged conduct ran from June 13, 2023, to February 7, 2024. Like Company claims this breached Directive 2001/29/EC (the InfoSoc Directive), the press publishers’ right under Directive (EU) 2019/790 (the Digital Single Market (DSM) Directive), and Hungarian copyright law. It contends that, when prompted, Gemini generated summaries echoing substantial parts of its articles, which it says shows the underlying large language model was trained on that content.

Google disputes that Gemini’s outputs amount to a reproduction or a communication to the public under EU law. It argues that its responses do not reach a “new public” beyond the internet users who could already access the original articles. It also says Gemini is a generative tool rather than a content database. In Google’s account, Gemini can reproduce specific text only if a user’s prompt explicitly references the underlying article.

On April 3, 2025, the Budapest court stayed the domestic proceedings. It referred four questions to the Court of Justice of the European Union (CJEU) under Article 267 of the Treaty on the Functioning of the European Union. The questions ask whether large language model training constitutes reproduction under the InfoSoc Directive. They also ask whether the text and data mining exception in Article 4 of the DSM Directive can apply, and whether chatbot output constitutes a communication to the public.

On March 10, 2026, the CJEU Grand Chamber heard roughly six hours of argument on generative artificial intelligence and copyright, with the European Commission and several member states intervening. The Commission argued that some questions focused on facts rather than law and were partly or wholly inadmissible. Member states divided over the territorial reach of EU copyright law where AI training occurs outside the European Union.

Advocate General Maciej Szpunar did not deliver his non-binding opinion on its scheduled date of September 3, 2026, and delivered it on November 3, 2026. He advised the Court to declare the two questions on large language model training inadmissible as hypothetical. He also advised that a chatbot’s real-time retrieval and reuse of press publications to generate answers may, in certain circumstances, infringe press publishers’ rights. No judgment date has been set, and the case remains pending before the CJEU.