A groundbreaking interdisciplinary study titled “Copyright & Training of Generative AI – Technological and Legal Foundations,” presented on September 4th at the European Parliament, reveals that the training of generative AI models involves copyright infringement. Commissioned by the Copyright Initiative, the study was conducted by Prof. Dr. Tim W. Dornis (University of Hannover) and Prof. Dr. Sebastian Stober (University of Magdeburg).
The research sheds light on the intermediate steps involved in AI training, revealing that this process does not qualify as text and data mining under German and European copyright law. Instead, it constitutes copyright infringement, as parts of the training data can be memorised and reproduced by AI models with the right prompts.
Axel Voss, MEP, who hosted the event, praised the study for offering vital insights into balancing copyright protection with AI innovation. Legal experts and representatives from the European Federation of Journalists (EFJ) and the Copyright Initiative emphasized that this study provides the technological and legal foundation for addressing the large-scale unauthorized use of intellectual property in AI training.
The study marks a significant step towards protecting creators and establishing fair licensing models as generative AI continues to grow.
For more details on the study, visit urheber.info.
