Sony Music removes 135,000 deepfake songs from music streaming platforms

Sony Music has removed 135,000 deepfake songs from several streaming services, including Spotify and Apple Music.

The songs were created using generative AI and targeted some of the company's biggest acts, such as Beyoncé, Queen, and Harry Styles.

Streaming fraud, also known as streaming manipulation, involves "fake" artists uploading song to sites like Spotify, YouTube, and Instagram, and artificially boosting their views in order to gain royalty payments.

The figure of 135,000 was disclosed at the release of the Global Music Report by the International Federation of the Phonographic Industry (IFPI) last Wednesday, March 18.

'AI slop' was reported to be often shared around the same time the targeted artist try to promote new music.

"That is when deepfakes are at their worst - building off and benefiting from the demand the artist has created (and) ultimately detracting from what the artist is trying to accomplish," said president of Sony's global digital business Dennis Kooker in a statement issued to the BBC.

The 135,000 removed AI-identified tracks are only a portion of AI-generated content uploaded across music streaming platforms.

Due to fast development of inexpensive and more accessible AI softwares, generative AI has become a prominent tool used in the music industry. Reportedly, since March of last year, Sony Music flagged around 60,000 songs falsely claiming to be from its pool of artists.

The French company Deezer was the first streaming giant to roll out its own AI-detection tool. The software labels songs that are 100% AI-generated, automatically excluding them from your playlists. Although Deezer flags AI-generated songs, it doesn't take them off the streaming service altogether.

Similarly to Spotify, which actively encourages users to report AI slop, Deezer also acknowledges the distinction between songs that are 100% AI-generated and songs where a fraction has been AI-generated for 'creative purposes'.

Apple Music uses 'transparency tags', which ultimately leave responsibility of disclosing AI content to labels and distributors, giving them the option to reject them entirely.

While Spotify introduced its impersonation policy last year, explaining how it assesses AI vocal clones, the platform still hasn’t introduced AI content tags

The release of the Global Music Report coincided with last week's publication of a UK government's AI regulation report. The law was supposed to enable AI companies to use copyrighted works to train their models.

The currently backtracked proposal had received major backlash from several artists, including Elton John and Dua Lipa.

Elsewhere, many writers and journalists, including Irish Emma Donoghue and Marian Keyes took part in the empty book protest against the use of their work to train AI. You can read more here.