My works appear in an AI training dataset. What is BumaStemra doing about it?

We are aware of the reports about AI training datasets and the tools that allow creators to check whether their works are included. We understand the concerns this raises.

In 2024, BumaStemra exercised its opt-out right under Dutch copyright law. This means that anyone wishing to use the repertoire we represent for text and data mining to develop AI systems must first obtain permission and provide fair compensation to rights holders. Without that permission, the use of your work constitutes copyright infringement and provides the legal basis for claiming compensation.

Our approach
Several court cases are currently underway in Europe concerning the use of copyrighted works for AI training. For example, our sister societies GEMA and Koda have initiated legal proceedings against major AI providers.

These cases are expected to provide greater clarity on questions that are important to our members. Can AI companies use copyrighted works to train their systems? And if so, under what conditions?

As these rules are largely determined at the European level, the outcomes of these cases will also have important implications for the Netherlands. BumaStemra supports these legal proceedings and closely monitors developments.

BumaStemra is committed to:

• Agreements with AI providers on permission, licensing and fair remuneration
• Clear and enforceable rules on which works have been used for AI training
• Legislation that protects the position of creators
• Supporting legal proceedings that can help provide legal clarity