Sony can now measure how much foreign music is in a track

Sony has developed a new technology that can detect traces of other artists’ songs within a composition and quantify it as a percentage – essentially measuring how much material from existing tracks has made its way into a generative music piece. If proven effective in real-world use, this tool could give rights holders a powerful way to claim royalties and audit who and how their libraries are being used to train AI models.

How Sony’s technology works

The system compares finished compositions against a database of existing music and delivers a “percentage of borrowing.” According to Nikkei, with the consent of AI developers, Sony can integrate directly with generative models; without that consent, the tech will still analyze the final track and match it to reference material. Sony says the tool isn’t just limited to music-it can also be applied to video, characters, and game content.

This isn’t some magic detector that instantly flags everything “smelling like AI.” Instead, it’s an evolution of audio fingerprinting tech, like YouTube’s Content ID or services from Audible Magic and ACRCloud, but focused on quantifying how much existing recordings contribute to a generated output. The key difference: Sony positions its system as a way to measure the extent of borrowing, rather than just a blocking tool.

The winners and losers here are clear: major labels and rights holders, including Sony’s own divisions, gain a significant leverage point to negotiate payments and enforce claims. Streaming platforms and generative music startups will face more legal scrutiny and commercial pressure. Independent developers and users fall into a gray area of uncertain rules and potential takedowns.

Meanwhile, listeners remain confused. A Deezer survey finds that 97% of users can’t distinguish AI-generated tracks from “live” recordings, underscoring the importance of detection technology for transparency. The big question is whether Sony’s tool will become a standard for verification and monetization or evolve into a licensing service for the industry. How fast the business moves from disputes over training data to revenue sharing depends heavily on that.