Prozesse: EuGH-Gutachten: KI-Chatbots können Urheberrecht verletzen

The European Court of Justice (EuGH) has ruled that generative AI chatbots like Google's Gemini may infringe copyright laws. According to the court's assessment, if these chatbots retrieve and utilize press articles from websites to provide responses to users, this could potentially violate the rights of publishers to reproduce and make their content publicly accessible.

This case marks the first instance of generative AI being addressed by the EuGH in Luxembourg. The opinion is not binding on the judges, and a final judgment is expected at a later date.

The general counsel, Maciej Szpunar, outlined the criteria for determining copyright infringement. The infringement depends on whether the chatbot unlawfully accessed the article, potentially bypassing technical measures such as a paywall. Szpunar emphasized that the rights of publishers do not protect the reported facts but rather the "specific expression" of the publications.

Publishers should not have a monopoly on news. The key point is whether the response uses the expression in a recognizable way, such that readers no longer have a reason to consult the original text. The public accessibility right is generally violated if the expression in the responses is "systematically recognizable."

This legal dispute originated in Hungary, where the press agency Like Company sued Google for allegedly unlawfully replicating and publicly distributing protected texts without permission. The Hungarian court had forwarded questions to the EuGH regarding EU law. Ultimately, the Hungarian court must decide the case, adhering to the guidelines set by the EuGH.

Regarding AI training, two questions posed by the Hungarian court were deemed inadmissible by Szpunar, as there was no evidence that the publisher's articles were used in training the Gemini models. However, he noted that the use of protected works for training an AI model is generally permissible under a legal exception known as text and data mining.

This allows automated reading and analysis of texts and data to detect specific patterns, provided certain conditions are met. Google stated that it welcomes the opinion on AI training. Gemini learns linguistic patterns to generate original content rather than reproduce existing content. Requiring parties would also have the option to object to the use of their content.