Musicians and AI: Warner Music and STIM Reinvent Compensation

A New Era for Musician Compensation

In the world of music, artists have always been compensated for the use of their works, whether through physical sales, online streaming, or royalties for covers and other uses. This economic model is based on a simple logic: the more a work is used, the more it generates revenue for its creator. However, the emergence of generative AI has disrupted this dynamic by redefining what it truly means to "use" a musical work.

Using a work to train an AI model could be perceived as a one-time event, occurring only during the initial training. However, creators might legitimately feel that the essence of their work continues to live on through the outputs generated by the AI, each time it produces a new result. This complexity has prompted companies like Sureel and SoundVerse to explore new ways to compensate artists in the age of AI, seeking to transform an industry often criticized for its appropriation of copyright into a fairer model.

Reinventing Copyright in the Age of AI

Sureel, a startup recently integrated into Warner Music Group, has partnered with the Swedish copyright agency STIM. Their goal is to find ways to compensate artists when their works are used to train generative AI tools. Sureel has developed innovative software that tags online music files according to the owner's instructions. These instructions determine whether an AI company can freely use the file for training, limit it within a training set, or exclude it entirely.

Sureel's software then tracks the use of these files by AI companies and establishes licensing fees accordingly. Meanwhile, SoundVerse rejects the idea of a one-time purchase of copyright, which it deems insufficient, as they wrote in a 2025 working paper. The company advocates for ongoing participation of artists in the AI lifecycle, emphasizing that each time an AI system generates an output, certain training data play a more crucial role than others. Thus, jazz music might have a more significant influence on a result than folk music, for example, justifying differentiated compensation.

The generative AI industry is often accused of being "the largest copyright theft in history." This criticism highlights the need to reform how creators are integrated into the AI development process.

The Challenges of Attribution of Influence

Attributing the influence of a work on an AI-generated output goes beyond simply measuring similarity between training data and the output. The real challenge lies in establishing a causal relationship between the training data and the AI, explains Tamay Aykut, CEO of Sureel. Even if the industry were to overcome this obstacle, it could encourage some to create music specifically to maximize copyright, a phenomenon already observed with streaming platforms that have influenced song structures.

To address this issue, Aykut proposes using advanced principles of information theory or modeling the actual historical impact of individual works. In a well-designed attribution system, atypical musical works could even have intrinsic value greater than radio standards. Simon Gozzi, head of business development at STIM, mentions that the company is exploring how Sureel's attribution reports could serve as a basis for licensing agreements between musicians and AI companies.

Public opinion fears that generative AI threatens cultural vitality by concentrating power in the hands of tech companies, devaluing creative workers, reducing revenues in the creative sector, and flooding the internet with low-quality content.

Towards Tailored National Policies

For Benji Rogers, co-chair of Sureel, attribution must be "multi-layered and auditable," requiring expertise in computer science, musicology, law, and economics. Governments wishing to remain competitive in the field of AI will need to support institutions capable of developing these policies.

Even the most liberal economies recognize the need to support cultural expression, whether through public funding or local broadcast quotas. As the impact of generative AI on the creative sector becomes increasingly evident, measures such as taxing large AI companies and redistributing those revenues to creators could prove essential in ensuring positive social outcomes. This approach could represent a new form of AI attribution, aligned with the interests of creative workers.