
The Hamburg District Court has rejected a photographer's complaint against the use of his work by an AI, citing the exception for text and data mining (TDM) for research purposes in Article 3 of Directive DAMUN 2019/790. The decision is not definitive.
When this Copyright directive 2019/790was adopted, there were no LLMs or large language models such as ChatGPT, nor multimodal LLMs that could be used to switch between text, a prompt and images. It was therefore not designed for such use. This new decision raises two important questions:
Should the training of generative AI models be regarded as text and data mining under §44b UrhG (German copyright law, equivalent to Article L122-5-3 of the French Intellectual Property Code)?
What is a research organisation? Who can benefit from the legal exception and be exempted from copyright reservations in order to make free use of protected works? What happens if the beneficiary then commercialises the results?
In this case, the photographer's website contained a provision in its general terms and conditions prohibiting data mining. However, the court contrasted this restriction with Article 60d UrhG (Article 3 Dir. 2019/790) cited above, which allows it to be circumvented if data mining is carried out for scientific research purposes. It considers that the subsequent commercial use of the training data is not an obstacle if the initial reproduction is carried out by an association.
The Hamburg court ruled that the reproduction of images for the training of artificial intelligence is covered by Article 44b UrhG (Article 4 Dir. 2019/790), without definitively resolving the doctrinal debate on its applicability to the creation of training models for artificial intelligence. Although Directive 2019/790 is silent on the subject, the Court considers that this interpretation is in line with recent technological and legal advances. It relies on Article 53(1)(c) of the European AI Act, which requires providers of general purpose AI to comply with the reservations set out in Article 4 of Directive 2019/790.
This decision sets a very low bar for copyright protection, given the exponential development of training data in recent years. We need legal clarification on the scope of the TDM exceptions in the context of new technologies! I'm curious to see how the French judge would interpret this.
Credits : Photo by Vera
Comments