German Court Rules in Favor of AI Company LAION in Copyright Case (2.10.24)

Key Highlights:

Court’s Interpretation of Section 60(d) of German Copyright Law: The Hamburg Regional Court dismissed photographer Robert Kneschke’s copyright infringement claims, citing Section 60(d), which allows text and data mining for scientific research purposes. The court ruled that LAION’s use of Kneschke’s images to train AI was protected by this exception.
Impact on Copyright Control: The ruling has raised concerns among creators, as Section 60(d) limits their ability to control the use of their work for AI training, even if they explicitly opt-out. The court highlighted that the “reservation of rights” mechanism, often relied on to prevent unauthorized use, does not apply in cases of scientific research.
Potential Precedent for AI and Copyright Law: This decision is considered surprising by many copyright experts and could set a significant precedent in the EU regarding the use of copyrighted materials for AI training. It also raises questions about how the upcoming EU AI Act will intersect with copyright laws in the future.

In a decision that has raised eyebrows across the creative and legal sectors, a German court recently ruled in favor of LAION (Large-scale Artificial Intelligence Open Network) in a copyright lawsuit filed by photographer Robert Kneschke (Kneschke v LAION). The case, the subject matter is the use of Kneschke’s images in the LAION 5B dataset to train AI models without his consent, hinged on whether LAION’s activities fell under specific exceptions in German copyright law. The Hamburg Regional Court ruled that LAION’s use of Kneschke’s images was permissible under Section 60(d) of the German Copyright Act, which provides a copyright exception for text and data mining (TDM) for scientific research. This decision has sparked widespread debate on the balance between AI development, intellectual property rights, and copyright law.

The Lawsuit: Copyright Infringement Allegations

Robert Kneschke, a stock photographer, filed a lawsuit against LAION in April 2023, claiming that his images were used in the LAION 5B dataset for AI training without his permission. LAION’s dataset is freely accessible and has been used to train popular AI models like Stable Diffusion. Kneschke alleged that this unauthorized use constituted copyright infringement, as the images were sourced from stock photo websites where automated use, including AI training, was explicitly prohibited in the terms of service. LAION, a nonprofit organization, defended its actions by invoking exceptions to copyright infringement under Sections 44(a), 44(b), and 60(d) of the German Copyright Act, which allow for certain uses of copyrighted works without permission.

The Court’s Decision: Section 60(d) Exception for Text and Data Mining

The court dismissed Kneschke’s lawsuit, finding that LAION’s use of his images for AI training fell under Section 60(d) of the German Copyright Act. This provision, implementing Article 3 of the EU Copyright Directive, permits text and data mining for scientific research purposes. The ruling emphasized that LAION, as a nonprofit organization, met the requirements for this exception, as its activities were aimed at scientific research and the datasets were made available for free.

“The fact that the dataset in dispute may also be used by commercially active companies for training or further developing their AI systems is, however, irrelevant to the classification of the defendant’s activities,” the court stated. LAION’s association with commercially driven AI models was deemed insufficient to invalidate its nonprofit status or the classification of its work as scientific research.

The Impact of Section 60(d) on Copyright Protections

“Text and data mining for scientific research purposes

(1) It is permitted to make reproductions to carry out text and data mining (…) for scientific research purposes in accordance with the following provisions.

(2) Research organisations are authorised to make reproductions.

ʻResearch organisationsʼ means universities, research institutes and other establishments conducting scientific research if they

pursue non-commercial purposes,
reinvest all their profits in scientific research or
act in the public interest based on a state-approved mandate. (…)”

This ruling surprised many copyright experts because Section 60(d) limits the control creators have over their works when it comes to scientific research purposes. Even though EU law allows creators to reserve their rights through a machine-readable opt-out mechanism under Section 44(b), this reservation of rights was found to be inapplicable when Section 60(d) is invoked.

“Section 60(d) limits the control creators have over their works and therefore many photographers and others may find it troubling that Kneschke was unable to prevent the use of his photo,” said Ronak Kalhor-Witzel, counsel at Norton Rose Fulbright in Munich. She highlighted that this ruling could be concerning for creators who rely on the reservation of rights as a means to control the use of their work.

The Photographer’s Response and Potential Appeal

In his response, Kneschke expressed surprise at the court’s decision, particularly regarding the reliance on Section 60(d). “At the conciliation hearing, we were informed that more detailed submissions would be required in this regard and that a decision on this could only be made at a later date if necessary,” Kneschke noted. He added that the court seemed to change its approach unexpectedly, leading to the dismissal of his case. Kneschke has indicated that he is considering an appeal, seeking to challenge the ruling and clarify whether creators can retain more control over the use of their work in AI training datasets.

Implications for AI Development and Copyright Law

This ruling could have far-reaching implications for AI developers and copyright holders across Europe. By relying on Section 60(d), the court has essentially expanded the scope of AI training activities that can be considered scientific research, even if the datasets are eventually used by commercial entities. The case also raises important questions about the future of copyright protections in the EU, particularly as the EU AI Act approaches.

Many legal experts are now questioning how the upcoming EU AI Act will intersect with copyright law. Kalhor-Witzel suggested that this ruling could set a precedent for how AI training is regulated in the EU, potentially limiting the ability of creators to prevent the use of their work in training datasets. “The ruling could affect the upcoming EU AI Act and raise questions about how copyright protections will be enforced in the context of AI training,” she said.

Conclusion: A Precedent-Setting Ruling?

While the court’s decision in favor of LAION has been described as “legally sound,” it has also created a sense of unease among creators who fear that their rights may be diminished in the face of rapidly advancing AI technologies. Kneschke’s potential appeal will likely be closely watched, as it could either affirm or challenge the growing trend of allowing broader use of copyrighted works for AI training under scientific research exceptions. The case has highlighted the need for clarity on how copyright law will adapt to the age of AI, as creators and developers alike navigate the evolving legal landscape.