An adviser to the Court of Justice of the European Union said Tuesday that AI chatbots that fetch and reuse news articles in real time to answer users may, in certain circumstances, infringe press publishers’ rights, while training AI models on protected works is generally allowed. The opinion, in the first generative AI case to reach the EU’s top court, sketches the liability tests that Google and other AI providers serving European users may have to design around.
What Happened
Advocate General Maciej Szpunar delivered his opinion in Case C-250/25, Like Company v Google, according to the court’s press release. Like Company, a Hungarian operator of web portals, sued Google in the Budapest Regional Court. It claims Google used its articles to train the models behind its Gemini chatbot and repeatedly reproduced and made them available through Gemini between June 13, 2023, and May 31, 2024. The Hungarian court asked the EU court four questions.
The case turns on retrieval-augmented generation, a technique in which a chatbot looks up current web pages, such as news stories, and draws on them to write its answer. Szpunar said a publisher’s right is not infringed simply because a chatbot reports facts. The right, created by the EU’s 2019 copyright directive, protects how an article is written, not the news itself, and is not meant to give publishers a monopoly over the news.
He proposed two tests for the reproduction right. The first is whether the chatbot reached the article unlawfully, for example by getting around a paywall or anti-scraping measures. The second is whether the answer reuses the article’s wording so recognizably that a reader no longer needs the original. The provider is liable where such use is inherent to its business model or where it fails to put credible, effective countermeasures in place, the opinion says.
On training, Szpunar advised the court to declare the two questions inadmissible as hypothetical, because there is no evidence the articles were used to train Gemini and training is not “online use” of press publications. He added that training on protected works is, as a rule, permitted under the EU’s text-and-data-mining exception.
Why It Matters
The opinion is not binding, but the court’s judges usually follow their advisers, Reuters reported. If they do, AI providers in the EU would face less exposure over training data and more over what their products show users. Any company offering an AI assistant or AI search tool in Europe that draws on news content would need to show it does not get around access controls and that its answers summarize facts rather than stand in for the article.
Reuters noted that publishers such as Like Company have seen traffic fall as AI-generated news summaries replace search results with links. A ruling against Google could force it to pay fees for such content, Reuters reported.
What to Watch
The judges have begun deliberating and will rule later; the court gave no date. The Budapest court will then apply the ruling to the facts.
