The music industry's real AI threat isn't what you can hear; it's what you can't see. A new generation of tools is being engineered not just to create AI music, but to make it indistinguishable from human work and impossible to trace.
We spoke with Virginie Berger, the Founder of advisory firm Don't Believe the Hype and a rights expert with leadership experience at industry cornerstones like Songtrust and MatchTune. Her message is stark: while the industry fights over yesterday's problems, a more sophisticated and elusive threat is already here.
Dark arts: “Forget about simple AI detection. We're now dealing with what I call a ‘darknet of copyright laundering,’” Berger says. “There are entire services that exist just to ‘humanize’ AI-generated music, making it indistinguishable from human-created work.” The real problem, she warns, isn't just that AI models are being trained on unlicensed tracks; it’s that a sophisticated, covert ecosystem is being built to erase every trace.
Hiding in plain sight: "An AI engineer showed me how easily you can poison a track to delete the pattern that detection tools look for," Berger says. "Or for £150, a service will re-record your AI track with real voices and instruments. If someone’s paying for that, they’re trying to hide." This deliberate laundering, she argues, makes proposals like an “AI Music” chart meaningless.
The deception isn’t accidental; it’s engineered. And it exposes the flaw in calling artists anti-innovation. “GenAI is a different beast because it scraped music without any consent, monetization, or authorization to create systems that directly compete with the original work.”
Divided they fall: As AI surges ahead, the music industry is too splintered to hold the line. “It was so strange to see headlines about Suno and Udio talking with record labels. They can’t do anything without the publishers and songwriters. You can’t just license a recording without the underlying composition,” Berger states. That weakness is compounded by distributors like Spotify, who welcome AI because they stand to benefit from a flood of royalty-free content.
Pot, meet kettle: Berger isn’t buying the “fair use” defense from tech giants who claim that paying for training data would kill innovation. “Look at the hypocrisy. Tech leaders want to erase intellectual property, but infringe on one of their logos and they’ll sue you in a heartbeat.” She says these companies won’t come to the table unless they’re forced, and that starts with a major legal win. “In Europe, we have hubs like ICE for streamlined digital licensing. We need a ‘Merlin for AI’—a unified front that includes majors, indies, and publishers—to negotiate with real leverage.”
Caught in the crossfire: “The independent creators are the ones getting screwed here, completely,” Berger says. “They were already squeezed by the DSPs, and now they are doubly impacted because they don’t have a major label to protect them.”