rr
The growing use of artificial intelligence (AI) in various creative industries, particularly photography, has brought new possibilities but also significant challenges, especially regarding copyright. A key case currently unfolding in Hamburg, Germany, highlights these challenges.
Photographer Robert Kneschke has filed a lawsuit against LAION e.V., a German organization that manages one of the largest datasets used to train AI image generators like Stable Diffusion. The dataset, known as "LAION 5B," contains over five billion images paired with text descriptions, collected from the internet without seeking consent from the original creators.
Kneschke discovered that some of his copyrighted works were included in the LAION 5B dataset. He argues that this violates his rights under German copyright law, as his images used without his permission to train AI models. Represented by copyright lawyer Sebastian Deubelli, Kneschke is challenging LAION e.V. in court, claiming that the organization engaged in unlawful reproduction of his works.
The case has gained significant attention as it has the potential to set a precedent for using copyrighted material in AI training. LAION e.V., however, argues that their actions are legally permissible under specific exceptions in German copyright law.
Photo by Steve Johnson from Pexels
One of the central legal arguments revolves around whether LAION e.V.’s activities can be classified as "text and data mining" under § 44b of the German Copyright Act (UrhG). This provision allows for automated analysis of digital works to extract information, as long as the original works are legally accessible and not explicitly excluded by the rights holder.
The organization contends that they only temporarily stored the images during the data-gathering process, which was automated using web crawlers, and that this is covered under these exceptions. However, in Kneschke’s case, the image rights were managed by a stock photography agency that had explicitly prohibited automated data collection in its terms of service, raising questions about the validity of LAION's defense.
Another key issue is the concept of "machine-readability" of copyright reservations. LAION e.V. claims that only a clear exclusion of specific web crawlers in a robots.txt file qualifies as machine-readable. In contrast, Kneschke’s legal team argues that the stock agency's natural language prohibition on automated scraping should suffice.
The case also touches on the broader implications of AI in creative industries, as it questions whether the large-scale use of copyrighted material for AI training should be allowed, even if such use is claimed to be for scientific research. LAION e.V. maintains that their activities fall under scientific research exemptions, but Kneschke argues that the organization’s close ties with commercial entities like Stability AI undermine this claim.
Photo by Jaiju Jacob from Pexels
On July 11, 2024, the Hamburg Regional Court held its first hearing in a case that had been delayed for several months. The court acknowledged the significance of the case, given its potential impact on artists and technology companies alike. While LAION e.V. does not dispute that it downloaded the images in question, it argues that this action was still legally justified. The court is carefully examining whether the exceptions cited by LAION apply and whether Kneschke’s claims hold up under the current law.
As the legal battle unfolds, the court is considering whether to refer the case to the European Court of Justice (ECJ), given that the relevant laws are based on EU directives. However, for now, the Hamburg court has opted to handle the case directly, with a ruling expected on September 27, 2024.
In a ruling on Friday, the Hamburg Regional Court determined that LAION (Large-scale AI Open Network) benefits from an exception under Section 60(d) of Germany's copyright law. This section allows privileged research organizations to reproduce works, even if the author has opted out. The court found that LAION qualifies for this exception because it releases its datasets for free, classifying its activities as non-commercial and for scientific research.
The court further clarified that the dataset's use by commercial companies to train or develop AI systems is irrelevant to LAION's classification as a research entity. Additionally, the fact that some LAION members engage in paid work for these companies does not link their commercial activities to LAION.
Norton Rose Fulbright counsel Ronak Kalhor-Witzel expressed surprise at the ruling, noting that the judge's decision did not consider other sections of the German Copyright Act that address valid opt-out provisions. Stock photographer Kneschke, who brought the lawsuit, called the judgment "strange" and is contemplating an appeal.
The ruling could have broader implications for the creative industry, particularly photographers and other creators concerned about the limits of control over their works.
Such cases prove that having a strong online presence as a photographer increases your exposure and visibility, but it also heightens the risk of your photos being stolen and used inappropriately. Therefore, understanding photography copyright and knowing how to protect your image content online is crucial.
Do you find this article helpful? Share it with friends on social media.