Court confirms legal use of images for AI training

A year and a half ago we had two interviews in DOCMA 106: with Robert Kneschke and the lawyer for the LAION association. Kneschke is a photographer and saw his copyrights violated by the unsolicited use of his images to train image-generative AI – the association that creates such data sets considered the use to be entirely legal. An initial verdict has now been made: the use of the images for AI training is legally permissible.

The court has opened the cage door and images available on the web can be used as training data sets for image-generative AI. Good news for AI users – some photographers will regret the court’s ruling. | Illustration: Doc Baumann with ideogram

In a case that is important for the open research and further development of artificial intelligence (AI) in Germany and Europe, on September 27, 2024, the Hamburg Regional Court dismissed the lawsuit brought by a stock photo dealer against the non-profit association LAION eV in all points (ref.: 310 O 227/23). The court thus confirmed the legality of the defendant’s actions in the creation of training data sets and rejected the allegations of the plaintiff, who had claimed a violation of his copyrights. The association was represented by the law firm Heidrich Rechtsanwälte from Hanover, which specializes in AI law.

The focus of the legal dispute was the use of an image of the plaintiff as part of the creation of the publicly available data set “LAION 5B”, which is used for training generative AI models in the image area. The plaintiff saw the one-time, temporary, automated reproduction of one of his images as a violation of his copyrights. Contrary to what is often portrayed, the data set does not contain any images, but only links to images that can be found elsewhere on the Internet.

LAION eV, which operates as a non-profit organization and is dedicated to promoting scientific research, argued, however, that reproduction was permissible within the framework of the text and data mining process (TDM) and was covered by the statutory exceptions to copyright law, in particular Section 60d of the Copyright Act. Lawyer Nick Akinci, who represented the association, particularly points out the importance of classifying AI training as TDM: “This corresponds to the clear will of both the European and German legislators”.

The court’s decision

In its ruling, the Hamburg Regional Court did not see any inadmissible reproduction of the copyrighted image. Instead, the court confirmed that the defendant’s actions were consistent with an exception provision (so-called limitation rule) of copyright law that was specifically created for scientific research purposes. This meant that the reproduction for the purposes of AI research by the non-profit organization was justified and legal.

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.

Significance of the judgment for AI development

This ruling represents an important milestone for research in the field of generative artificial intelligence, as it confirms the legal use of copyrighted images for training data sets such as “LAION 5B”. Since its founding, LAION eV has been committed to promoting research and providing free, public data sets that are available to the scientific community worldwide.

“We welcome the court’s decision, which underlines the importance of free and public data sets for the research and development of AI models,” explained Christoph Schuhmann and Dr. Jenia Jitsev as representative of LAION eV “The ruling gives us legal certainty and enables us to continue to promote innovative and scientifically based projects with a high degree of transparency and reproducibility that benefit society as a whole.”

LAION eV will continue to support research in the field of artificial intelligence in accordance with its purpose and provide free data sets for scientific purposes. The judgment of the Hamburg Regional Court creates an important basis for the legally secure use of publicly accessible data in the context of scientific research. It confirms that the association can continue to make a significant contribution to promoting open source initiatives in the future, which will particularly promote AI development in Germany.

The plaintiff can now appeal the judgment to the Hanseatic Higher Regional Court within one month.

Full text of the judgment at openjur.de

For further information please contact:

Heidrich Lawyers
Attorney Nick Akinci
Tel.: 0511 – 37498150
[email protected]