Spotify finally appears to be taking the problem of AI slop seriously – it’s announced that it has removed more than 75 million tracks that it has identified as slop or spam.
In addition to this, a further 50 million tracks have been filtered out of recommendations for users after repeated complaints about their quality. In a newsletter that it has posted online, Spotify said that: “We looked closely at the tracks listeners were skipping, ignoring, or telling us they didn’t want to hear. A clear pattern emerged: these tracks were consistently ones identified as slop or spam.”
It comes a year after the platform unveiled a series of measures to stop AI slop from overwhelming their system, including a new spam filtering system to identify uploaders and stop the tracks from being recommended by its algorithm. It also cracked down on ‘impersonation violations’ ie fraudsters who create deepfakes of an artist to hoover up royalties.
In addition to the filtering system, the platform has also put in place a ‘verified by Spotify’ scheme which shows listeners that there is a real human being behind a particular track – over 300,000 artists are now verified on Spotify.
In the newsletter, Spotify pointed out that it has used a combination of tools to identify AI slop: “listener engagement and feedback, upload patterns, metadata, and information from multiple third-party AI detection tools.”
However, they admit that as generative AI technology develops it will inevitably get harder. So they are looking at new tactics, including “mass uploads, duplicates, SEO manipulation, spammy playlist networks, and artificially short tracks. As spammers change tactics, we’ll keep updating our systems.”
It is a constant battle, which the platform acknowledges. But after a number of years during which Spotify seemed quite content to let AI completely swamp their system, their belated decision to combat the problem appears to be a step in the right direction on at least one front.
