29
votes
Spotify removed 75m spam tracks over the past year as artificial intelligence tools increase the ability of fraudsters to create fake music
Link information
This data is scraped automatically and may be incorrect.
- Authors
- Dan Milmo
- Published
- Sep 25 2025
- Word count
- 600 words
I recommend Venus Theory's video explaining the different types of fraud targeting musicians lately.
https://youtube.com/watch?v=plleJ0Zv0Ww
Most of it comes down to music distribution industry practices having evolved in a higher trust and lower volume environment, which revolves around honesty in metadata, and doesn't have hardening to deal with democratized access. So people can apply false track metadata, which some distributors don't catch, and it affects what streaming services show users.
Ah, this explains some weird tracks I’ve noticed under artists I enjoy but clearly aren’t their music, especially on platforms that are not Spotify.
I assumed it was just a case of shared artist name and immature database practices, but track metadata fraud makes way more sense.
Verification does not seem that hard of a problem to solve, luckily.
I'm glad to see that Spotify understands that royalties going to ai-slop undermines their business model of licensing music by human artists. The phrase "fake music" is quite funny though.
In this case, it's not a qualitative judgment over the music, but fraud. Things like releasing a track with metadata matching a known artist, so streaming services automatically list it under that artist, but they collect royalties. AI music generation has just enabled fraudsters to churn out audio files that pass the minimal scrutiny much faster. So they impersonate artists, or build playlists full of 31-second junk audio that gets played by bot farms to rack up plays.
Yes I understand what the scam is. It's a pretty boring "generate some stuff, lie about what it is" scam that snake-oil salesmen have been doing for ages.
What's interesting to me is to consider what "fake music" would mean in a different context. As someone who listens to experimental stuff I could see applying it, tongue-in-cheek, to some genres I enjoy. And at a more "theory of art" level... could we call random noise fake music, or is it real music? When does the intent of the artist make irregular clanging into music and what would make it real?
It's AI slop. Which is to say, low-effort fully automated machine output devoid of creativity or attention to detail. Still technically music, but bottom of the barrel. Worse than stereotypical canned Muzak elevator music.
I don't know where one would draw the line between noise and music. That's an older question that predates generative AI. I've seen a single piano note preceded by 15 minutes of recorded silence, passed off as a postmodern musical composition. I've heard recordings of dissonant dot matrix printers grinding their way through MIDI arrangements, which are at least technically interesting but sound awful. Far be it from me to gatekeep what qualifies as art but I can at least distinguish between what I like and what I don't.
I'm actually a big supporter of generative AI as a useful tool for artists (as opposed to a replacement for them). I've spent a lot of time working with the Stable Diffusion ecosystem, and probably even longer fiddling with Udio, a generative music platform. Both will happily produce slop if you just throw some word salad into the prompt box and hit submit. But where it gets interesting, for me, is in running many concurrent generations and curating the outputs for quality, and iteratively refining what remains by inpainting away the slop, bit by bit. Also including more traditional tools in the workflow... I'll use Photoshop to manually retouch my images, and I love running Udio tracks through a DAW to clean them up and properly master the final result.
I think the existence of a spectrum from "one-click crappy art machine" to "novel tool for an artist's toolbox" is lost on most people debating the value of AI art. These opportunistic Spotify spammers are clearly just cranking out huge volumes of garbage. At the same time, I'm in the process of composing an album using AI tools, and I'd love to distribute it on the streaming platforms. I'd hope the effort and intent behind mine would be apparent and I wouldn't be reflexively categorized as just another peddler of "fake music."
Side note: I fully expect many well-known musicians have already integrated AI into their workflows. But because they're professionals who know the ins and outs of music and production, they're not releasing lazy slop and they don't want to call attention to their use of it. I think generative AI is going to become one of those "little secrets" that everyone in the industry uses but no one talks about, like airbrushing was for photography, autotune was for music, and VFX and digital beauty work still are for film.
That's cool you don't like AI music. I have similar concerns but notice I didn't say anything about AI?
These are the concepts I'm interested in. Atonal Set Theory is an area of research in music theory that I've followed a bit. I don't think there is a particularly interesting discussion to have, but turning the phrase "fake music" over in my mind is fun, like a mental fidget spinner.