The European Artificial Intelligence Act (AIA) and the 2019 Directive on Copyright in the Digital Single Market (CDSMD) provide the legal framework for Dutch remuneration claims. The CDSMD grants rightsholders the option to prohibit the use of their works for AI training purposes, while the AIA obliges AI developers to provide summaries of works used for AI training. Together, these regulations empower rightsholders to control their works’ use, but it is unclear how fair remuneration and appropriate AI revenue distribution could be achieved under these frameworks. Against this background, we are conducting a reality check to clarify how these EU regulations could be implemented to ensure fair remuneration for artists and rightsholders.

• WP1 (Computational musicology) is assessing the extent to which current AI models incorporate specific copyrighted works. We are examining how traces of the training data appear within music embeddings (the ‘mind’ of the models) and are regenerated in AI compositions.
• WP2 (Sociology) is examining how artists perceive AI’s role in music creation, whether it affects their income or offers new opportunities, and whether remuneration is needed for works used in AI training (including willingness to pay when using AI tools themselves). Surveys and interviews are gathering data on artist experiences, income impacts, and compensation expectations.
• WP3 (Law) is determining how the CDSMD/AIA regulations can be applied effectively (and potentially also improved) in the light of insights from WP2, identifying opportunities and challenges for securing remuneration for artists and rightsholders. This package is also proposing fair revenue partitioning schemes based on insights from music matching in WP1.

This research translates musicological and sociological insights into legal recommendations. An interdisciplinary and multi-method approach, integrating explainable AI (XAI) methods, a survey and semi-structured interviews, and legal analysis is crucial for fully understanding the nuanced and complex challenges of remuneration. This project will be among the first to:

• quantify the influence of specific training examples on GenAI music compositions, an-alysing their encoding paths within the ‘brain’ of GenAI systems;
• investigate how GenAI affects working conditions for artists and rightsholders, exam-ining their responses to GenAI opportunities and challenges in the context of income and creative control; and
• develop guidelines for fair remuneration and licensing, balancing the needs of artists, rightsholders, and AI developers by combining insights from computational, musico-logical, sociological, and legal perspectives.