Music streamer Spotify released its latest quarterly results on Tuesday, reporting 300 million premium subscribers and 14% revenue growth. The company also pointed to rising costs tied to AI-powered features. The music industry is wrestling with the best way to incorporate and acknowledge artificial intelligence in the business. AI tools are one thing, but what about AI-generated music or AI-generated artists?
Last summer, a rock band called The Velvet Sundown reached viral streaming success on Spotify with their song “Dust on the Wind.” Both the band and the song were generated by AI.
Now, a year later, record labels are proposing new guidelines on how music generated by artificial intelligence should be treated. To discuss, “Marketplace Morning Report” host Kimberly Adams was joined by Dan Runcie, founder of the music and entertainment research group Trapital. The following is an edited transcript of their conversation.
Kimberly Adams: So, what are the record labels proposing in terms of how the industry should be handling music made by AI?
Dan Runcie: So, last week Wednesday, there was this coalition led by three of the biggest music companies in the world: Universal Music Group, Sony, and Warner, and a few others as well. And they proposed that any music recording that involves the use of generative AI, — and if it needs to qualify for any charts — has to meet three conditions. First, the service has to use properly authorized and lawful music. Second, it needs to be substantially human-made. And then third, the activity can’t raise chart manipulations or any type of fraud concerns.
Adams: Tell me more about this idea of blocking songs that have concerns around raising stream or chart-manipulation concerns.
Runcie: The fact that AI has now increased the amount of music that is uploaded to these streaming platforms considerably, it now increases the risk, and we’ve seen that reflected in the data. There’s one streaming company called Deezer, and Deezer had reported recently that 90,000 tracks that are uploaded to its service daily, nearly half of the tracks that it has, are AI — generative AI music. And a majority of those were fraudulent tracks. And I’ve heard enough from others in the industry that if you can assume that these facts are true about Deezer, then that likely is similar to some of the other streaming platforms that we’ve seen. To be clear, a lot of these tracks aren’t being listened to, but these are all measures to try to fix that problem.
Adams: Yeah, I imagine it makes it harder to find real music by real humans.
Runcie: It does. And if I could say — because this does tie into that second point, the “substantially human-made” — this has been one of the pain points, because what counts as “substantially human-made” is a very gray area. And I think it was intentionally left gray with these measures, because on one hand they’re still trying to figure it out, and it’s hard to be precise when we’re building the plane as we’re flying it. But still, this represents how tangled this challenge really is.
Adams: So, the record labels have made this request/demand about how AI music should be handled when it comes to charts or considerations. But who actually gets to decide whether or not these are the rules that we’re going to follow?
Runcie: It goes through a few stages. At least what we’ve seen publicly so far, the record labels have this coalition that put it forward. There’s also one of the influential bodies in the industry called the IFPI, and they recently had adopted these rules as well. But I think what you would specifically wait to see is the charts themselves. Charts matter a lot to this industry. Record labels and the artists themselves put a considerable amount of marketing dollars to have a No. 1, or to have a No. 2, or to make sure that they’re on the top of the charts. So anything that is a threat is something that they do put focus on. And Billboard Charts, for instance, which of course is powered by the Luminate data, if we see these charts, then adopting those principles moving forward, that’s where it really impacts things, at least for that particular aspect.
