I welcome the decision by some music platforms to label fully AI generated music.
There is a real problem. A person can generate thousands of tracks in seconds, upload them in bulk and collect tiny payments that become profitable only because almost no time, labour or thought was invested.
Deezer now receives almost 75,000 fully AI generated tracks every day, representing 44 per cent of all new uploads. It also reports that 85 per cent of the streams generated by such tracks in 2025 were fraudulent. Label them. Remove manipulated streams. Keep automated spam out of recommendations. That makes sense.
But then comes the much more dangerous term: “AI assisted music”.
What does that even mean?
A composer may write every note, every chord and every word, then use AI to change the instrumentation. A synthesizer once allowed us to turn a keyboard into strings, brass or sounds that had never existed before. Today AI can help us move between instruments and voices with even greater freedom.
A composer can upload their own voice and use it to sing what they cannot physically perform. Is that fundamentally different from pitch correction?
One experienced producer has estimated that 99 per cent of modern pop uses corrective Auto Tune. It is an industry estimate, not a scientific measurement, but few producers would dispute that pitch correction has become close to universal. A doctoral study also found that 93 per cent of its respondents used digital audio workstations. Another study of 2,399 Billboard songs found drums in 96 per cent and synthesizers in 29 per cent.
If we start excluding music because technology helped realise it, there will not be much music left.
Perhaps one per cent.
Come to think of it, that is roughly the number of artists who already seem invited to share most of the money. A UK Competition and Markets Authority study found that more than 60 per cent of streams went to music from just 0.4 per cent of artists.
But jokes aside, AI is also part of a democratisation of music creation.
A person who cannot play an orchestra can still compose for one. A creator with a physical disability can turn an idea into sound without first mastering an instrument their body may not allow them to play. Someone without money for musicians, studios and engineers can hear an arrangement, test an instrument or receive advice while remaining the person who wrote the music and the lyrics.
Technology can remove the barrier between imagination and sound. It does not automatically remove the human being.
Of course, the industry must distinguish this from automated music factories. But today’s detection technology cannot do that with certainty. Deezer has acknowledged that detection rates can fall sharply when systems encounter new models or unfamiliar data. Researchers also warn that AI music detectors can produce both false positives and false negatives and should not be used alone for decisions affecting artists or distribution.
We have already seen music penalised for alleged artificial streaming even when artists say they neither ordered nor knew about it. Independent creators are often left with little evidence and limited ability to defend themselves.
Do we really believe the same system will give a red card to the world’s largest artists if AI has been used somewhere in their production?
Probably not. The favourite team, forgive me, the favourite artist, will most likely keep the royalties and collect another medal.
So yes, label fully AI generated mass production.
But do not confuse the tool with the creator.
And do not reduce a human composer’s value because technology helped their music travel the final distance from imagination to sound.