Gemini: chatbots could replace articles – publishers’ rights are at risk

A chatbot can turn an article into such a comprehensive response that there is no need to visit the website where it was published. And, in certain circumstances, it may infringe the publisher’s rights. This is the line taken by the Advocate General of the Court of Justice of the European Union, Maciej Szpunar, in the first European case to address the relationship between generative artificial intelligence and publishers’ rights.

The dispute pits Google Gemini against Like Company, a Hungarian firm that operates online portals. Like Company claims that Google used its articles to train the chatbot’s language models and that Gemini reproduced them on several occasions in its responses to users, making them available between 13 June 2023 and 31 May 2024.

The case is before a Hungarian court, which has asked the Court of Justice of the European Union to clarify what protections publishers are entitled to.

On the first point, Szpunar recommends that the claims relating to training be declared inadmissible: he notes that there is no evidence in the case file that Like Company’s articles were used to train Gemini. However, he adds a comment on the merits: as a general rule, the use of protected material to train a model falls within the European exception for text and data mining.

The key issue, however, concerns the extraction of articles from the web and their reuse in responses. Szpunar points out that the protection afforded to publishers does not extend to facts and news as such, but to the expressive form of the publications and the investment required to produce them. To assess a possible infringement, two aspects must be considered: whether the chatbot accessed the content unlawfully – for example, by circumventing a paywall – and whether the response recognisably reproduces the form of the article, thereby relieving the user of the need to consult the original publication.