‘AI false positive becoming more common in music’

Artificial intelligence – referred to as AI – and its implications have been dominating the chat in the entertainment industry, whether it’s Hollywood or the global music industry. Recently, dancehall veteran, Papa Michigan, took to social media to “challenge AI” after one of his songs was flagged as AI by music distributor, TuneCore.

Michigan, in the social media post, showed the studio and the musicians who played the live music on the track, and emphasised that he did not use AI at any stage of the production.

He told The Gleaner that he was shocked when he received emails from TuneCore, that a song on his So Many Roads album, was not eligible for distribution.

“After reviewing your release, our content review team identified indicators suggesting the use of generative AI tools that rely on datasets that are not properly licensed, which does not align with TuneCore’s requirements. This applies to the following tracks: Track 7 - Who Can Be Against Me. As a result, this release is not eligible for distribution at this time,” the letter said in part.

Michigan submitted a formal appeal, in which he emphasised that “this track is a 100-per-cent human studio recording” and requested a review with a human moderator. He removed the song and released So Many Roads on August 14. The album is available on all major streaming platforms under Rashanco Music.

General manager of the Jamaica Music Society (JAMMS), Evon Mullings, told The Gleaner that “the scenario of some human-made music being rejected (false positive) by digital distributors is becoming increasingly common”.

“Digital distributors such as TuneCore, DistroKid, CD Baby, and others, are coming under increasing pressure from the major streaming services such as AppleMusic, Spotify and Deezer Music, to function as the gatekeepers in tagging music that are AI-generated or AI-assisted, before the music reaches the retail platforms,” Mullings explained.

He added, “As a result, the digital distribution platforms have been implementing aggressive, automated AI-detection software to screen uploads. DistroKid for example is being sued by Universal Music Group after being accused of uploading thousands of ‘music made with AI’ without labelling them as such. They are being accused of deceptive in their business practices and so these distributors, even prior to the lawsuit, have been implementing very strong AI-identification measures.”

He shared that a producer, or self-producing artiste, whose human-created music is labelled as ‘made with AI’ can file an appeal to the digital distributor’s support team and submit “proof of creation”. This would include “video evidence, high resolution DAW screenshots, or offer to send individual raw audio stems”.

“Secondly, human music can sometimes accidentally trigger AI filters because of how it was mixed. As such, the producer should check if every single drum hit is perfectly locked to the grid (100 per cent quantised), and if the vocals are heavily processed with aggressive Auto-Tune. Under these conditions, detection algorithms can mistake the lack of ‘human imperfection’ for robotic generation and flag it as music made with AI,” he added.

LABELLED ‘MADE WITH AI’

He stated that streaming services are not outright banning or rejecting ‘music made with AI’, they want the music it to be labelled as such – ‘Full AI-Generated Music’ or ‘AI-Assisted Music’.

“The streaming services are using different methods to identify ‘music made with AI’,” Mullings explained.

AppleMusic and YouTube, he said, are relying on mandatory disclosure by the uploaders through labelling or tagging.

“Others, such as Deezer Music, rely on their own in-house AI detection tools to automatically tag fully AI-Generated music for transparency purposes. Spotify relies on voluntary disclosure and uploader-supplied metadata tags rather than universal tagging,” Mullings explained.

The JAMMS general manager stated that the streaming services need for full AI transparency is directly related to the consumer “so that those who do not wish to purchase ‘music made with AI’ can make that choice”, and also the competition factor.

“Record producers, record labels, songwriters, composers, artistes, and musicians, and their industry bodies have also been lobbying heavily for the human-interest to be protected. ‘Music made with AI’ is competing with music made from human creativity and ultimately displacing the potential of creators to earn fully from their works,” Mullings, who is an attorney-at-law, told The Gleaner.

He stated that many streaming services are excluding ‘music made with AI’ from editorial playlists, algorithmic recommendations, and official charts.

“Spotify applies an “AI Persona” label to separate pure AI artiste profiles, isolating them so they do not show up in regular user recommendations. These measures are helping to limit the impact that the vast presence of AI music is having in the global music market,” he said.

Addressing how music ‘made with AI’ affect monetisation and royalty earnings, Mullings explained that in standard music streaming, royalties are paid out based on a track’s percentage of total platform streams.

“Because millions of AI tracks are being uploaded and streamed by fraudsters using bots, they are effectively stealing money out of the pool meant for human artistes.

Mullings said that JAMMS, in keeping with best practices emerging globally, which are designed to better manage music rights in the age of AI, is introducing revised requirements around its ‘song registration’ process.

yasmine.peru@gleanerjm.com