Almost 40% of everything released in July shows traces of a machine: a detector checked a million records
23.2% of July releases were fully generated, another 15.3% were machine audio a human then processed. And a third of the artists whose music tested positive said they had used nothing of the kind.
Almost 40% of the music released worldwide last month shows signs of a machine's involvement. More than a million July releases were run through SubmitHub's SH Labs detector: 23.2% came back fully generated, and another 15.3% were machine-made audio that a human had "modified or processed".
The study was covered by Mixmag.
SubmitHub unveiled the detector in June, framing the job not as a hunt for AI but as a response to a "growing disclosure problem".
"I think the path forward for AI music is disclosure. People should be able to decide for themselves whether they want to engage with AI-generated music. Our job is to give them enough information to make that choice," said SubmitHub founder Jason Grishkoff.
A third say there was no AI
The study's most uncomfortable number is not about machines but about people. Among artists whose releases tested positive this year, 31% told the platform, when asked directly, that they had used no such tools at all.
That is the "disclosure problem" in one line: the question is not whether you may make music with a machine, but whether you will say so. Until someone says so, the choice Grishkoff wants to leave to the listener has nothing to work with.
Who already uses the detector
SH Labs is used by Bandcamp and Traxsource, and by the German collecting society GEMA — the same body that won its case against the generator Suno in July. A tool born inside a song-pitching service has become part of the infrastructure that stores and rightsholders now lean on.
As reported earlier by ONE//FM, a Munich court found that Suno had no right to train its models on music represented by GEMA.
Read also
Grammys pull HUGEL’s hit from consideration over AI. He asks them to show where the machine is
Two seconds of someone else's rhythm and courts running since 2004: how the landmark sampling case ended
Spaces between the letters slip past the filter, and the machine sings in a familiar voice: four musicians are suing Suno
First they cleared their own chart, then they turned to radio: Australia is the first to tell the machine its place
"Made With AI": Apple Music makes the machine-origin label mandatory for labels
"Too good to live on a screen": Suno will sell vinyl of music its own algorithm generated
AI in music: what is already banned, what gets labelled and who is counting it
Six DJs in ten will not play it: Beatport has banned fully AI-generated tracks
Comments
Loading…