Court rejects plain language clause for AI training
Hamburg's Court of Appeal finds a plain language reservation of use insufficient for images used in AI training. The photographer lost the case for the second time.
For the second time, stock photographer Robert Kneschke has lost a legal battle regarding the use of his images for AI training. The Hamburg Court of Appeal largely upheld the lower court's decision, ruling that the image data provider Laion did not violate Kneschke's rights when it used his photos to train artificial intelligence models.
The core issue revolved around copyright law's exceptions for text and data mining (TDM). The court determined that Kneschke's reservation of use, stated in natural language within source code and website terms, was not sufficient to restrict AI training under German law. Specifically, the court noted that for commercial use under §44b, a machine-readable reservation is required, and for scientific use under §60d, no reservation is possible.
Judges concluded that while AI crawlers can compare images and descriptions, this does not imply they can process extensive terms of service. The court also affirmed that Kneschke is not entitled to remuneration for this use, aligning with the legislative decision to exempt certain TDM activities from royalty obligations.
This ruling is significant as the EU is reviewing its Digital Single Market Directive, which governs TDM exceptions. The European Commission is expected to present a report on the directive's implementation by June 2026. The Hamburg court has granted leave to appeal to the Federal Court of Justice, acknowledging the fundamental legal questions raised by AI and copyright.