Laion versus stock photographer: court allows data sets

The Hamburg Regional Court has dismissed a lawsuit brought by a stock photographer against the non-profit association Laion. Laion offers training datasets that can be used to train AI models. For this purpose, the provider creates link lists including the corresponding data. The stock photographer had filed a lawsuit, claiming that the procedure infringed his copyright. Laion is represented by the law firm Heidrich Rechtsanwälte, Joerg Heidrich is legal counsel for heise Medien.

The court confirms the legality of Laion's action. The freely accessible data set "Laion 5B" contains an image of the photographer. Or rather, it contains a link that leads to the image on the Internet. There is also a description of the image and information in text form about what can be seen in the image. Laion has checked the images itself using software to ensure that the content and text match. For this purpose, Laion downloaded the image, analyzed it and, according to its own statements, deleted it again.

The photographer's image was provided with a watermark and contained a note stating that it could not be reproduced. The lawsuit was based on this. The question was therefore whether the creation of the data records by Laion constituted a lawful reproduction or not. Heidrich Rechtsanwälte wrote in a press release: "According to the court –, the reproduction of an image for the purpose of selecting training data constitutes text and data mining within the meaning of the statutory exceptions." –

Processing of the image for research permitted

The grounds for the judgment also state specifically that although Laion had reproduced the photograph, this was covered by Section 60d of the copyright law. The paragraph is an exception for research purposes. This means that the processing of the image was permitted for research purposes.

According to the court, there is no need to conclusively assess whether the limitation provision of Section 44b of the Copyright Act applies against this background – "Whether the defendant can invoke the limitation provision of Section 44b of the Copyright Act appears doubtful". Section 44b regulates text and data mining.

The question of the legality of the use of Laion's data set for the training of AI models by AI providers themselves is also not answered by the ruling. In this context, however, the judge refers to a study by the Initiative Urheberrecht. It also states that when the training dataset was created, it was not foreseeable what would happen downstream – during training and when generating images.

In the grounds for the ruling, the judge also refers to European copyright law (DSM). The AI Regulation states that providers of AI models with a general purpose are obliged to find strategies to comply with applicable copyright laws.

The representatives of Laion e.V. welcomed the ruling in a press release: "The ruling gives us legal certainty and enables us to continue to promote innovative and scientifically sound projects with a high degree of transparency and reproducibility that benefit society as a whole." The association will continue to provide free data sets for research in the field of artificial intelligence.

The plaintiff has one month to appeal.

Disclaimer: The law firm of Heise lawyer Joerg Heidrich is representing LAION in the proceedings. He was not involved in the creation of this article.

(emw)