The market for AI music tools is expected to increase more than tenfold in the next five years, generating $3 billion in revenues by 2028, according to a new report commissioned by collecting societies GEMA and SACEM. As a result, 27% of music creators will be at risk of losing revenue to generative AI by 2028, which means - while many artists and songwriters are already using AI tools themselves - they also want more protections to be put in place.
The rapid development of generative AI, says GEMA CEO Tobias Holzmüller, "offers enormous economic potential. The works created by [songwriters] are the basis for this revolution. Nevertheless, from the perspective of many music creators, the risks so far outweigh the opportunities. This can only change if we now shape the overall conditions in such a way that everyone involved can participate appropriately in the success of this development".
The report was produced by research company Goldmedia. Citing stats from Market.us and S&P, it estimates that the global market for generative AI - so AI models that generate content - currently brings in around $3.7 billion in revenues.
Of that, approximately 8% comes from music AI platforms, so $300 million. With the estimated ten fold increase to $3 billion by 2028, in a few years time music AI revenues will equal 28% of all the money collected by the music industry's song right collecting societies in 2022.
Artists and songwriters themselves are among those making use of music AI tools. In a survey of 15,000 songwriter and publisher members of GEMA and SESAC, 35% have used AI technology in some way as part of their creative process. Use of such tools differs according to age and genre. 51% of respondents under 35 have made use of AI - and creators making electronic music are most likely to use AI tools, with 54% saying they had already done so.
At the same time, the majority of music creators fear that ever more sophisticated generative AI is ultimately a threat, with 71% admitting they are "afraid that the use of AI for music could lead to music creators no longer being able to make a living from their work".
The report considers which areas of music-making are most threatened by AI. It notes that "companies or businesses looking for more generic music to soundtrack their assets or commercials may already be satisfied with AI-generated music in the future". And within the music industry, artists and labels will likely rely more on AI tools for backing tracks and mastering, inevitably reducing opportunities for session musicians and studio producers.
For the music industry at large, the threats are balanced by opportunities. Although the size of the opportunity does depend to an extent on how AI companies choose to work with the music community. This in turn brings us to the big debate over the copyright obligations of AI companies and whether an AI model can be trained with existing music without getting permission from - and therefore negotiating licences with - the relevant copyright owners.
Unsurprisingly, the vast majority of those surveyed - 90% - believe permission must be sought, and that creators and copyright owners should benefit financially if their music is used to train AI. 95% also said that AI companies should declare what content has been use to train their models and 89% stated that AI-generated tracks should be identified as such.
"The figures in the study show that the estimated damage could be considerable for creators", SACEM CEO Cécile Rap-Veber states. Urging the tech sector to work with the music industry, she adds "we hope to establish a transparent and fair relationship between creators and AI companies".
Meanwhile, on transparency, she references the relevant provisions in the European Union AI Act, and concludes, "we call on the French and German governments not to oppose the implementation of effective transparency requirements for generative AI companies".