The regional court dismissed the plaintiff's claim, although it acknowledged that the reproduction of the photo did, in principle, infringe on the plaintiff's copyright. However, it held that this use was covered by the text- and data mining (TDM) exception for scientific purposes under Section 60d of the German Copyright Act (UrhG), as the defendant used the material for non-commercial scientific research purposes. The dataset is made freely available to the public and is intended to benefit other researchers for training AI models.
Section 60d of the German Copyright Act contains an exception to copyright law for the purpose of collecting and evaluating content in the interest of science. The purpose of the exception is to enable legally secure research without having to take into account the copyrights of third parties. The fact that the data set could later be used to train AI applications did not, in the court's view, mean that the exception was not applicable.
The court explicitly emphasized that this decision was not about the permission to use the image for training AI, but rather about the prior question of whether the defendant was allowed to download the image to compare it with his image description database.
The court then also expressed its legal opinion on another question, which was not relevant for the decision due to the applicable exception of Section 60d. This concerns the applicability of a second exception for text and data mining, namely Section 44b. The exception allows reproductions of “lawfully accessible” works for text and data mining. A work is “lawfully accessible” if it is freely available on the internet or if the user himself is authorized to access it.
Whether or not this section applied in this case would have depended on whether the terms of use of the photo agency's website had excluded its applicability. The requirement for this would be that the terms are “machine-readable”. According to the court, the definition if this term depends on the technical developments. In view of the progress in AI development, which now enables programs to capture the content of texts written in natural language, not only terms stored in the website code are covered, but also those formulated in natural language.