To judge from the results of a report commissioned by GEMA and SACEM, the specter of artificial intelligence (AI) is haunting Europe.
A full 35% of members of the respective German and French collective management societies surveyed said they had used some kind of AI technology in their work with music, according to a Goldmedia report shared in a Tuesday (Jan. 30) press conference — but 71% were afraid that the technology would make it hard for them to earn a living. That means that some creators who are using the technology fear it, too.
The report, which involved expert interviews as well as an online survey, valued the market for generative AI music applications at $300 million last year – 8% of the total market for generative AI. By 2028, though, that market could be worth $3.1 billion. That same year, 27% of creator revenues – or $950 million – would be at risk, in large part due to AI-created music replacing that made by humans.
Although many of us think of the music business as being one where fans make deliberate choices of what to listen to – either by streaming or purchasing music – collecting societies take in a fair amount of revenue from music used in films and TV shows, in advertising, and in restaurants and stores. So even if generative AI technology isn’t developed enough to write a pop song, it could still cost the music business money – and creators part or even all of their livelihood.
“So far,” as the report points out, “there is no remuneration system that closes the AI-generated financial gap for creators.” Although some superstars are looking to license the rights to their voices, there is a lack of legal clarity in many jurisdictions about under what conditions a generative AI can use copyrighted material for training purposes. (In the United States, this is a question of fair use, a legal doctrine that doesn’t exist in the same form in France or Germany.) Assuming that music used to train an AI would need to be licensed, however, raises other questions, such as how many times and how that would pay.
Unsurprisingly, the vast majority of songwriters want credit and transparency: 95% want AI companies to disclose which copyrighted works they used for training purposes and 89% want companies to disclose which works are generated by AI. Additionally, 90% believe they should be asked for permission before their work is used for training purposes and the same amount want to benefit financially. A full 90% want policymakers to pay more attention to issues around AI and copyright.
The report further breaks down how the creators interviewed feel about using AI. In addition to the 35% who use the technology, 13% are potential users, 26% would rather not use it and 19% would refuse. Of those who use the technology already, 54% work on electronic music, 53% work on “urban/rap,” 52% on advertising music, 47% on “music library” and 46% on “audiovisual industry.”
These statistics underscore that AI isn’t a technology that’s coming to music – it’s one that’s here now. That means that policymakers looking to regulate this technology need to act soon.
The report also shows that smart regulation could resolve the debate between the benefits and drawbacks of AI. Creators are clearly using it productively, but more still fear it: 64% think the risks outweigh the opportunities, while just 11% thought the opposite. This is a familiar pattern with the music business, to which technologies are both dangerous and promising. Perhaps AI could end up being both.