In September 2024, GEMA became the first collecting society worldwide to introduce a licensing model for generative artificial intelligence (AI). The goal of this model is to ensure that music creators receive a fair share of royalties when their works are used for training AI systems, generating new AI songs, or as part of AI-generated music content.
AI as an innovative new form of music utilization
Through licensing AI models and systems trained on copyrighted musical works, GEMA is responding swift-ly and appropriately to the rapid pace of technological advancement. The ongoing systematic collection and evaluation of the global music repertoire represents an unprecedented use of copyrighted materials, leading to extensive claims for participation from rights holders.
Balancing the interests of services and creatives
The intersection of generative AI and copyright-protected works raises numerous legal issues that remain unresolved. Licensing is essential for providing clarity and reliability to the developers of foundational gen-erative AI models and the AI music tools built upon them. This licensing also ensures that creators receive a fair share of the economic benefits generated by these AI offerings, taking into account that these benefits stem from the creative efforts of the original authors.
One licensing model - two key components
Generative AI presents new opportunities for revenue generation. The foundation of corresponding business models is built on artistic, creative, and journalistic content. A fair remuneration model starts where value is created. This model should not only consider the training of an AI system but also the value generated by AI-created content in the marketplace. Additionally, it is important to take into account the competitive land-scape with human-created works, as the development of AI content has been made possible by these earlier creations.
GEMA's two components model is designed to ensure that music creators receive a fair share of the eco-nomic benefits generated by generative AI. These benefits arise in two main areas: first, with the AI providers themselves, and second, with the subsequent use of AI-generated music content, such as on other platforms or as background music.
The two components license model guarantees that rights holders receive a continuing share of all econom-ic benefits generated from the training and subsequent use of AI generated music content.
First component: Licensing generative AI service providers
The licensing model is initially connected to the training of AI models, which includes both pre-training and fine-tuning. It does not matter how or in what format the musical works are utilized for training.
The licensing approach provides a 30% share of all net income generated by the generative AI model or system of the provider. This means that creative professionals participate for example in subscription fees collected by the AI tool providers, as well as other revenues generated from the AI output on these plat-forms.
If there are no economic benefits, or only minimal ones, creative professionals must still receive an appro-priate share for the use of their works. Therefore, GEMA's licensing model includes a minimum royalty in addition to the standard royalty, which is common practice. To fairly reflect the extent of rights utilized, this minimum does relate to the amount of AI output produced.
When calculating the royalty amount, it is essential to consider that using original works of music for gener-ative AI represents the most intensive forms of use. The creative contributions of the original work's authors are particularly significant in this context.
The first component of this licensing model targets all providers of generative AI services operating within the German market. The license is not dependent on where the training takes place; instead, it is linked to the generation of output. This approach acknowledges recent scientific findings, suggesting that the train-ing data remains embedded in the trained models.
Second component: Subsequent use of AI generated music content
The licensing concept encompasses not only the training of AI models and systems but also all economic benefits that can arise from the subsequent use of AI-generated music content (e.g., as background music or on music platforms). This is important because the latter is largely based on the original musical works and contains protectable elements of the works used for training as latest studies show. With this context in mind, the second component provides users who wish to reuse and monetize AI-generated music content with the assurance that they are complying with copyright law and are not at risk of infringing on rights. In the future, rights holders will so receive an appropriate share of the additional income generated by AI-produced songs. This share must be at least equivalent to what would have been provided for purely hu-man-generated works.
Conclusion
GEMA's two component licensing model provides a reliable licensing framework for both the training of AI systems and the subsequent use of AI-generated content. This approach ensures that music creators receive an appropriate share of the economic benefits derived from the use of their artistic works during AI training and in subsequent utilization. As a result, GEMA supports technological advancement while protecting au-thors and their creative contributions.