CJEU Opinion on AI and Copyright Law

Advocate General Maciej Szpunar of the Court of Justice of the European Union (CJEU) issued an opinion on what may constitute an infringement of press publishers’ rights by artificial intelligence (AI) chatbots, in the first case involving generative AI to reach the CJEU.

The preliminary questions referred to the CJEU by a Hungarian court arose from a case involving Like Company, a Hungarian press publisher, and Google, concerning its general-purpose AI, Google Gemini. Like Company alleged that:

·         Google had used its articles to train Google Gemini;

·         Google Gemini uses copyrighted content from its websites without authorisation or compensation; and

·         The summaries provided by Google Gemini may serve as substitutes for news publishers’ articles, ultimately affecting their revenue and visibility.

In light of these allegations, the Budapest Regional Court referred the following questions to the CJEU for a preliminary ruling:

·         Does displaying text in AI chatbot responses that partially matches press content protected under Article 15 of the Copyright in the Digital Single Market (CDSM) Directive constitute a “communication to the public” under EU copyright law, and does the fact that the response is AI-generated affect this classification?

·         Does training an AI chatbot by observing and matching linguistic patterns constitute an act of “reproduction” under EU copyright law?

·         If AI training constitutes reproduction, does training on lawfully accessible works fall within the text and data mining exception in Article 4 of the CDSM Directive?

·         Does a chatbot provider perform an act of reproduction when the AI generates output containing press publishers’ content in response to a specific user prompt that matches or refers to that text?

Advocate General Szpunar stated that there was no evidence that Google had used Like Company’s articles to train Google Gemini. Advocate General Szpunar clarified that training AI models on copyrighted material is legally permitted under the exception for text and data mining.

However, where AI chatbots retrieve press publications, reuse them to generate answers and provide those answers to users, press publishers’ rights may be infringed. The Advocate General emphasised that protection extends to the specific form of expression used in a publication, rather than the facts it reports. The aim is not to grant press publishers a monopoly over the news itself, but to protect their investment in producing publications.

Furthermore, the Advocate General explained that determining whether an AI system’s use of a publication falls within the scope of the publisher’s reproduction right depends on two conditions:

1.      Whether the AI accessed the publication unlawfully, for example by bypassing a paywall or paid subscription designed to restrict access; and

2.      Whether the answer provided to the user merely reports the facts contained in the publication or reproduces its form of expression in a recognisable manner.

Advocate General Szpunar concluded that an AI chatbot provider, in this case Google, would be liable for such acts where they are inherent in the provider’s business model, or where the publisher has failed to take appropriate steps to counter them.

Conclusion:

If the CJEU adopts Advocate General Szpunar’s opinion, many AI chatbot providers will need to rethink and refine how their chatbots retrieve information and present it to users, while news publishers’ and writers’ copyrighted works will receive greater protection. This will be particularly relevant where publishers and writers use paid subscriptions, as many outlets already do, including El Mundo in Spain, Bild in Germany and the Times of Malta locally. This would help strike a balance between the efficiency of AI and the protection of writers’ distinctive forms of expression.