AI chatbots such as Google’s Gemini could infringe copyright when they retrieve and use online news articles to answer users, according to an opinion by European Court of Justice Advocate General Maciej Szpunar. Such use may affect publishers’ rights to reproduce and make their works available to the public, the court said in a statement summarising his conclusions.
The case is the court’s first concerning generative artificial intelligence. The Advocate General’s opinion is not binding on the judges, who will issue their ruling at a later date.
Szpunar said whether reproduction has occurred may depend on how the chatbot accessed an article, including whether it bypassed technical restrictions such as a paywall. He also stressed that publishers’ rights protect the specific expression of their articles, not the facts they report, and do not give publishers a monopoly over news.
A key question is whether a chatbot answer recognisably reuses an article’s expression in a way that leaves readers with no reason to visit the original. The right to make a work available to the public would generally be infringed if that expression appeared systematically in chatbot responses, according to the opinion.
The dispute began in Hungary, where Like Company, a publisher that operates several news websites, sued Google. It alleged that Gemini reproduced and made its protected texts available without permission. A court in Budapest referred questions about EU law to the ECJ; the Hungarian court will ultimately decide the case, following the European court’s interpretation.
Szpunar considered two questions from the Hungarian court about training AI models inadmissible because they were hypothetical: there was no evidence that the publisher’s articles had been used to train Gemini. He nevertheless noted that using protected works to train an AI model may be permitted under the EU’s text and data mining exception, which allows automated reading and analysis of texts and data under certain conditions.
Sources: FAZ.