AI a Copyright: studie Úřadu Evropské unie pro duševní vlastnictví

The European Union Intellectual Property Office has released a study examining the relationship between generative artificial intelligence and copyright from technical, legal, and economic perspectives.

The study thoroughly addresses the use of copyright-protected content as training data for generative artificial intelligence models, generating their outputs, the emerging market for content licensing, and the role of the Office and national agencies.

The study analyzes in detail how authors can reserve rights regarding data mining of their works, specifically concerning the use of data mining techniques during the training of artificial intelligence. This particularly involves the opt-out regime options as regulated by EU legislation and the Czech national copyright law. It emphasizes that the quality of input data is crucial for the quality of the results generated, especially in case of specialized models.

Effective copyright protection can thus give rise to an entirely new licensing market and support not only authors in protecting the results of their work but also developers in creating generative models with high-quality outputs.

The study further discusses the requirements of European regulation regarding the transparency of outputs, specifically the content generated by generative artificial intelligence. The Office highlights the importance of accurate and complete information about the origin of data, their potential protection, and permissible uses. It also underscores the importance of providing information on whether the content was created using artificial intelligence.

According to the Office, future discussions should focus on ways to protect outputs produced by artificial intelligence. While generative artificial intelligence does not currently fulfil the legislator's notion of an author as a natural person capable of creative activity, there may be legitimate interest in protecting the results of its activities in certain cases.

The study concludes that there is currently no effective solution for the interactions between artificial intelligence and copyright requirements, particularly regarding the reservation of rights and their overall protection. However, new mechanisms are emerging for copyright holders and developers that could facilitate the process for both parties. In the fall of 2025, a Copyright Knowledge Centre will be established by the European Union Intellectual Property Office to address these topics.

Authors:

Vladimír Petráček

Jakub Marko

Veronika Juhasová