The music industry has a new open secret. Producers are using it. Songwriters are using it. Some of the biggest names in the business are using it on records you’ve already heard. And almost none of them want to say so out loud.
Songwriter Michelle Lewis describes the current atmosphere among her peers as “don’t ask, don’t tell.” Suno CEO Mikey Shulman put it more bluntly, comparing AI adoption in music to Ozempic — everybody’s on it, nobody wants to admit it. Lumineers producer David Baron says there’s a real “social penalty” around being seen as an AI user, even as the tools become standard parts of the workflow. Teddy Swims found that out firsthand when he called the tools “truly amazing” at SXSW Sydney and walked straight into a fan backlash.
That tension is the defining feature of where the industry sits right now. The public resistance hasn’t disappeared — it just went quieter than it was a couple of years ago, when the AI Drake and Weeknd clone “Heart on My Sleeve” hit 8.5 million plays before getting pulled from every platform. The outrage was loud then. Now it’s been replaced by something closer to reluctant acceptance. “No one wants to be left behind, or come across as old-school,” Lewis says.
What’s Actually Happening in the Studio
Most professionals aren’t asking Suno to write their songs. What they’re doing is more specific and, in many ways, more significant. Stem separation — the ability to isolate vocals or instruments from an existing recording — has become one of the most widely used tools in production. Baron called it “phenomenal,” describing how he isolated a vocal and it sounded like it had been recorded alone in a pristine studio. Two or three years ago, that wasn’t possible.
Young Guru, Jay-Z’s longtime producer and engineer, says AI-generated samples have quietly taken over a significant portion of hip-hop production. His estimate is that more than half of sample-based hip-hop is now being made with AI-generated funk and soul, rather than licensing original music or hiring musicians. The prompts have gotten specific — producers aren’t just asking for “soulful 1960s music” anymore. They’re asking for music recorded at Motown with a particular writer, or Stax with a specific bassist. The outputs have gotten detailed enough to match.
Songwriters in Nashville and LA are using tools like Suno to generate fully arranged demos from lyrics and chords, which they then shop to artists and labels. Lauren Christy of the Matrix described getting a text from a major artist asking if she had any songs ready, being able to respond immediately with a demo — melodies, lyrics, and chords all hers — and having the artist commit on the spot. “I was like, ‘Whoa — that just saved me days.'”
Charlie Puth drew his own line. He’s used AI to mock up sounds while building ideas for his new album, testing whether he wants a choir before committing to booking one. But he’s firm about where it stops. Using generative AI to transform a finished production, he said, is “nauseating.”
The Part Nobody’s Accounting For
Every task AI handles is work someone isn’t getting paid for anymore. Demo musicians, assistant engineers, studio owners, production music composers — these are the first casualties, and it’s already showing up. According to a survey of over 1,100 producers and engineers by Sonarworks, seven out of ten respondents were at least occasionally experimenting with AI tools, and one in five were regular users. Baron pointed out that the mundane session work AI is replacing is exactly where the next generation of producers gets trained. “Our generation is going to go away eventually,” he said. “We need those 25-year-olds trained.”
Lewis was more direct about the damage in certain sectors. Production music — stock tracks licensed for TV, radio, and media — is, in her word, “toast.” Children’s animation isn’t far behind. Smaller production companies that need to cut costs are finding what they can get away with, and they’re getting away with a lot.
There are creative consequences too. Lewis described a writing partner who built an AI demo with a vocal that had no pauses for breath — technically unsingable by a human. Producer Nathan Chapman raised the issue of “demo-itis,” the tendency of artists to get attached to a demo and want to replicate it exactly. AI demos add a strange wrinkle to that problem because they can sound simultaneously polished and off in ways that are hard to name. “I haven’t heard a Suno demo yet that sounded bad good,” Chapman said. “They’re all just… good.”
Baron has started noticing AI’s sonic fingerprints — irregular time signatures, a “weird 2/4 bar that no one would write.” He told the story of a world-class drummer being asked to replicate an AI drum part. The drummer found it humiliating.
The Copyright Problem Nobody Has Solved
Underneath all of this is an unresolved legal foundation. The questions around what Suno and Udio were trained on, how copyrighted material is being recombined in their outputs, and how artists and songwriters get paid for any of it remain open. Warner Music Group settled with both Suno and Udio late last year and struck licensing deals with both platforms, but Sony and UMG are still fighting Suno in federal court. Meanwhile, UMG is separately pushing for a ruling that Anthropic’s use of copyrighted lyrics to train its Claude model constitutes clear infringement — a case that could set the standard for how AI training and copyright intersect across the board. Even producers who are enthusiastic about the tools say they’re holding back on using AI in released music until the legal picture clears. “I’m scared to use anything,” Christy said. “I would hate to have an AI detector detecting something.”
The policy layer isn’t settled either. The UK government recently reversed its position on allowing AI companies to train on copyrighted material without permission — a significant signal that governments are starting to push back on the idea that rights holders should have to opt out rather than opt in. And the feds already landed a guilty plea in an $8 million AI-driven streaming fraud case that showed exactly how quickly AI can be weaponized to exploit the royalty system when no one’s watching closely enough.
