Hungarian Dolphins: Europe’s First Landmark AI Copyright Case
A Hungarian publisher, Like Company, has initiated a significant legal battle against Google concerning its AI chatbot, Gemini, which summarized an article about Hungarian singer Kozsó and his dream of introducing sweetwater dolphins from the Amazon to Lake Balaton. The legal action arises from the fact that Like did not grant Google express permission to utilize its article, prompting the publisher to sue.
This case has been escalated to the European Union Court of Justice in Luxembourg, with a hearing scheduled for March 10. While this case is unlikely to finalize the ongoing conflict between AI developers and publishers, it represents the first major ruling that will clarify how European copyright laws pertain to AI.
Contrasting US and European Copyright Laws
In the United States, courts have already issued significant rulings involving Meta and Anthropic, exonerating these tech companies from copyright infringement. The US legal system permits fair use, allowing unlicensed snippets of copyrighted material to be used for commentary, criticism, or research on a case-by-case basis.
Conversely, the European Copyright Directive provides additional protections for copyright holders. While it permits commercial text and data mining in certain instances, rights-holders can opt out. The recent AI Act in the EU has expanded these opt-out options, particularly concerning the training of chatbots. However, significant legal questions remain about the boundaries of copyright infringement regarding AI models.
The Implications of the Like Case
As Europe grapples with balancing creator rights and technological advancement, many officials fear the continent is lagging in AI development. To avoid hindering progress, there have been considerations to delay some stringent “high-risk” AI regulations until 2027. The European Parliament is contemplating whether to mandate AI developers to license copyrighted content, while Henna Virkkunen, the European Commission Executive Vice-President, has argued that AI training benefits from the text and data mining exemption in both the Copyright Directive and the AI Act.
The court must now clarify whether training a generative AI model falls under the commercial “text and data mining” exception of these laws or if publishers should be compensated for their content used in training.
Key Questions in the Case
Central to this dispute are several critical questions: Did Google utilize the content of Like’s article to train its AI model? Did this result in an “unauthorized communication to the public”?
According to Like publishers, Google exceeded the allowances of the text data mining exemption by using their copyrighted material to train Gemini, subsequently disseminating it to the public without authorization. Google, however, contends that even if it summarized the Like article, such an action does not constitute communication to a “new public.” Furthermore, Google asserts that the chatbot’s summary did not reproduce any parts of the article beyond central facts and insists that Gemini was not trained on the article in excess of the allowed data mining exemption.
Conclusion and Future Outlook
While the court’s ruling could set far-reaching precedents, some analysts remain skeptical about whether this case is suitable for establishing a pivotal legal standard. Paul Keller from the Institute for Information Law suggests it is “highly improbable” that Gemini provided the summary based on training from the copyrighted article. Instead, it is more likely that the chatbot retrieved the article from the web in real time to generate the summary, given the short timeframe between the article’s publication and the prompt.
This perspective shifts the focus from the training methods of AI chatbots to whether the reproduction of a copyrighted text by a chatbot constitutes an “unauthorized communication to the public.” Experts like Dr. Andres Guadamuz of the University of Sussex argue that summarizing an article accessible to the public cannot be deemed a communication to a new public, as it is available to anyone on the Internet and primarily communicated privately to the individual user.
Establishing precedents for emerging technologies is inherently complex, with far-reaching implications. While this case may not resolve all ambiguities surrounding European copyright and AI, it holds the potential to provide crucial insights and answers. As noted by Martin Kremtscher from the Centre for Regulation of the Creative Economy, a flawed ruling could leave Europe with a convoluted regulatory environment for AI developers and insufficient protections for rights holders.
