Musicians Could Get Paid Each Time AI Uses Their Songs

Musicians might finally get paid fairly when AI companies train on their songs, thanks to innovative tracking technology that treats creative work like streaming royalties.

Companies like Sureel and SoundVerse are building systems that trace which songs influence each AI-generated track. Instead of buying music rights once and using them forever, AI companies would pay artists ongoing royalties based on how much their work shapes the final output.

Sureel, recently acquired by Warner Music Group, created software that labels music files with permissions set by artists. The technology tracks exactly how AI companies use each song during training and calculates fair licensing fees. Meanwhile, SoundVerse argues that if an AI system creates jazz music, the jazz songs in its training data deserve more payment than folk songs that barely influenced the result.

"Attribution isn't about recreating the old economics," says Sureel Co-President Benji Rogers. "It's about measuring, for the first time, the thing the old economics only approximated."

The approach mirrors how musicians already earn money from streaming services, radio play, and cover versions. The more something gets used, the more creators earn. That simple principle motivates artists to keep creating.

Simon Gozzi from Swedish copyright agency STIM says his organization is exploring how attribution reports could shape licensing agreements between musicians and AI companies. The agency has already partnered with Sureel to test the system.

The Ripple Effect

This technology could reshape how creative industries interact with artificial intelligence. Instead of the "biggest act of copyright theft in history," as some critics describe current AI training practices, companies could coexist peacefully with artists.

The system might even reward experimental and unusual music more than radio hits. Sureel CEO Tamay Aykut suggests that quirky, unpolished works could hold more value for training AI systems than overplayed standards.

Major music companies are already negotiating. Universal and Warner recently struck deals with AI companies to work together on training models with proper copyright consent. These private agreements signal a shift away from the lawsuits that dominated early AI development.

Challenges remain. Attribution systems need sophisticated technology to avoid being gamed by people creating music specifically designed to capture royalties. The solution requires measuring actual influence, not just surface-level similarity to AI outputs.

Some industry voices prefer simpler approaches. Drew Silverstein from SourceAudio advocates for straightforward negotiated agreements with fixed annual prices paid at the point of training.

But supporters see a rare opportunity to build fair systems before bad practices become permanent, protecting both cultural vibrancy and the livelihoods of creative workers worldwide.