Deezer has revealed that AI-generated music now accounts for more than 50% of new tracks uploaded to the streaming service. In June, it reports, an average of 90,000 AI-generated tracks were delivered each day.
Different music streaming services have different policies towards AI-generated music. In the main, where platforms have deleted AI-generated tracks in the past, it was because of evidence that streams of those tracks had been artificially manipulated. Deezer, however, is now going further.
Alongside today’s big stat announcement, the streaming service says, “Moving forward, Deezer will systematically take down AI tracks used for streaming fraud and those that haven’t been streamed for at least six months”.
Given the sheer quantity of AI-generated music, and assuming streaming service algorithms don’t boost that music, you have to think a large percentage of AI-generated tracks will get minimal streams.
Deezer CEO Alexis Lanternier adds, “We’ve been at the frontline of fighting fraud and reducing payment dilution related to AI music for almost two years. Now that half of all daily uploads are AI-generated tracks, we are taking additional steps to safeguard the rights of artists and songwriters, while maintaining focus on music that fans actually love”.
Deezer has been particularly vocal about AI-generated music in recent years, regularly bigging up its proprietary AI music detection technology, and being the most forthcoming of all the streaming services when it comes to facts and figures about AI music uploads.
AI-generated music poses a number of questions for streaming services: Do they accept it? Do they label it? Do they recommend, push and playlist it? Do they allocate royalties to it? There is currently no real industry consensus on any of those questions.
There is also a debate as to how much we can depend on automated detection tools like those developed by Deezer to spot AI-generated music. They typically rely on “fingerprinting” by detecting audio artefacts left in the waveform.
As AI platforms rapidly evolve and users of those platforms seek to avoid detection, there’s an ongoing digital ‘arms race’ between detection techniques and scripts and services designed to strip out or minimise those AI artefacts to reduce detection.
The music industry is also developing labelling schemes encouraging artists and labels to self-declare AI-generated music alongside any automated detection. Though at the moment those schemes are voluntary and it’s not clear how widely they will be embraced within the industry at large, let alone by the people and companies who are prolifically and primarily uploading AI-generated tracks.
Plus, of course, there is also the extra tricky question of where to draw the line between AI-generated and AI-assisted music, assuming any restrictions that may be placed on AI-generated music may not apply to tracks where human creators have used AI as part of their creative process.
Last month one of the other smaller music streaming services, Tidal, also went quite big on its AI music policies. It declared that “Tidal will accept AI-generated music”, but it will be labelled and held to “a higher standard of content integrity” when it comes to things like streaming fraud checks. Plus, perhaps most importantly of all, “AI-generated music will not be monetisable”.
It remains to be seen to what extent the industry does align on these things. However, Deezer boss Lanternier seems to be a fan of a more joined up approach.
In Deezer’s latest AI statement earlier today, he concludes, “We have reached a pivotal moment, but we have the technology to keep track of the development, and an ambition to do what’s right. If the industry can align and come together to create shared standards, we can build a healthy and sustainable music ecosystem, where great music wins”.