EU court adviser says Gemini answers may infringe publishers’ rights

An adviser to the European Union’s top court has said that AI chatbots may infringe press publishers’ rights when they retrieve articles in real time and reuse their protected expression in answers. The opinion, issued Tuesday in a Hungarian case against Google’s Gemini, draws a distinction between the lawful use of facts and potential reuse of the way a news report is written.

Advocate General Maciej Szpunar advised the Court of Justice of the European Union that a publisher’s rights can be infringed where a large-language-model chatbot retrieves a publication, uses it to generate an answer and provides that answer to a user. His opinion is non-binding, and the court will rule at a later date.

The case was brought by Like Company, a Hungarian website operator, before the Budapest Regional Court against Google Ireland. Like alleges that Google trained Gemini on articles published on its web portals and repeatedly reproduced and made them publicly available through Gemini between June 13, 2023, and May 31, 2024. The Hungarian court referred four questions to the CJEU.

Two questions concerned whether using press publications to train a large language model falls under publishers’ exclusive reproduction rights or the EU’s text-and-data-mining exception. Szpunar advised the court to find those questions inadmissible in this case, because there was no evidence Like’s articles had trained Gemini and model training was not an online use of a press publication.

But he added that, as a general rule, using protected material to train a large language model is freely permitted under the text-and-data-mining exception. That exception is set out in Article 4 of the EU’s 2019 Digital Single Market Directive, while the press-publisher right at issue is in Article 15(1).

The remaining questions addressed Gemini’s outputs. The Advocate General said infringement would depend partly on whether a chatbot had obtained an article unlawfully, including by circumventing effective measures such as paywalls or anti-scraping restrictions. It would also depend on whether an answer went beyond reporting facts and reused the publication’s distinctive expression so recognisably that users no longer needed to consult the original.

The opinion said press publishers do not own the facts in their reporting. Their right protects the particular expression of a publication and the investment involved in producing it, rather than creating a monopoly over news. Systematic reproduction of that expression in chatbot responses to related queries would, as a rule, infringe the right to make a publication available to the public.

The court adviser also said a chatbot provider may be liable where such use is inherent in its business model, or where it has failed to introduce credible and effective countermeasures. Gemini uses retrieval-augmented generation, or grounding, to collect information from the web in real time, and may retrieve or scrape online press articles when answering news-related questions.

Google said the EU’s text-and-data-mining exceptions were important to European competitiveness and AI development, and welcomed the recognition that public information may be used for AI training. The company said Gemini learns linguistic patterns to produce original outputs rather than replicate content, and that publishers have controls to opt out.

The dispute comes amid publishers’ complaints that AI-generated news summaries replace searches that previously sent readers to media sites. As covered in October, Google has separately begun piloting payments to publishers whose content makes a significant contribution to responses from Gemini and its search AI products.

The Like Company case is expected to produce the CJEU’s first EU-level ruling on generative AI and copyright, in an area where no dedicated EU copyright legislation for generative AI exists.