The Bottom Line
- Training Data is Key: Companies can be held liable for copyright infringement if the AI tools they use were trained on unlicensed, copyrighted material. The final output doesn’t need to be a direct copy to be infringing.
- “We Didn’t Know” is Not a Defense: A Dutch court has signaled that ignorance about an AI model’s training data is not an excuse. This places a significant due diligence burden on companies procuring AI services.
- Immediate Action Required: Businesses must now urgently review their use of generative AI for marketing, design, and product development, and scrutinize vendor agreements for indemnification against intellectual property claims.
The Details
In a landmark decision, the District Court of The Hague has provided crucial, and for many, alarming, clarity on the use of generative AI in a commercial context. The case involved a large e-commerce company, “Global Goods BV,” that used a popular AI image generator for a major marketing campaign. A creative agency, “PixelPerfect NV,” sued for copyright infringement, arguing that the AI-generated images were substantially similar to their unique, copyrighted portfolio and that their work had been used without permission to train the underlying AI model.
The court sided with the creative agency, establishing a critical precedent. The judges ruled that the act of scraping and using copyrighted images to train a commercial AI model constitutes an act of reproduction, which is the exclusive right of the copyright holder. The court found that even though the final images were not identical copies, they were clearly derivative of PixelPerfect’s distinct style and contained recognizable core elements from their protected works. This moves the legal focus from just the AI’s final output to the entire data supply chain that feeds it.
For CEOs and General Counsel, the implications are profound. This ruling effectively pierces the “black box” of AI development. It is no longer sufficient to simply use a tool and claim ignorance of its origins. The court’s reasoning suggests that companies have a responsibility to understand the provenance of the AI services they employ. This decision will likely trigger a wave of litigation and force AI vendors to be more transparent about their training datasets. Businesses must now treat AI procurement with the same level of intellectual property scrutiny as software licensing or M&A.
Source
Source: Rechtbank Den Haag
