Spotify has revealed it removed 75m spam tracks from its platform over the past year as artificial intelligence tools increase the ability of fraudsters to create fake music.
The world’s biggest music streaming service announced a crackdown on vexatious tracks after admitting the rise of powerful AI tools had coincided with a significant amount of spam content being tackled by the streamer.
AI-generated spam is becoming a problem for streaming platforms and musical artists because every play more than 30 seconds long generates a royalty for the scammer behind it – and denies payment to a legitimate artist.
The 75m spam tracks rival the scale of Spotify’s actual catalogue, which stands at 100m tracks. Spotify also offers nearly 7m podcasts and 350,000 audiobooks.
The company said the spam tracks were identified either before they were uploaded as part of an existing filtering process for new tracks, or taken down after getting on to the platform and being identified as illicit.
Spotify said it would start rolling out a music spam filter to identify uploaders, tag them and stop the tracks from being recommended by its algorithm. The company said AI tools had made it easier to generate spam content such as impersonations, ultra-short tracks and mass uploads of artificial music, which range from meditation instrumentals to duplicates of famous artists.
“Spam tactics … have become easier to exploit as AI tools make it easier for anyone to generate large volumes of music,” Spotify said.
The Stockholm-based company, which has nearly 700 million users, said despite the uptick in harmful uses of AI-made content, it was not having a serious impact on listening habits or payments to artists. Spotify paid $10bn (£7.4bn) in royalties last year, although the level of royalty payments is often a subject of tension between the platform and artists.
“Engagement with AI-generated music on our platform is minimal and isn’t impacting streams or revenue distribution for human artists in any meaningful way,” Spotify said.
In 2023 Spotify introduced a rule that individual tracks have to be streamed more than 1,000 times before generating a payment, a change the company says has helped tackle scammers.
Spotify is also strengthening rules on vocal deepfakes, which are allowed only when the artist being impersonated has given their permission. It is also cracking down on scammers uploading deepfake tracks to a popular artist’s profile page.
One of the most notorious musical deepfakes was published in 2023. Heart on My Sleeve, a song featuring AI-made vocals purporting to be Drake and the Weeknd, was pulled from streaming services after Universal Music Group, the record company for both artists, criticised the song for “infringing content created with generative AI”.
Spotify said it would support a new industry standard for disclosing the use of AI in creating a track, developed by a tech and music-industry backed non-profit called DDEX. Artists’ use of the standard on the platform will be voluntary, said Spotify, and they will not be forced to label music as entirely or partly AI created.
“This change is about strengthening trust across the platform,” said the company. “It’s not about punishing artists who use AI responsibly or down ranking tracks for disclosing information about how they were made.”
The popularity on Spotify of Velvet Sundown, an AI-generated band, had given rise to calls for mandatory tagging of music created by the technology. Spotify has not taken down Velvet Sundown – which says it is a “synthetic music project” on its Spotify profile page – because it does not violate the company’s anti-spam policies.