“If you don’t build this stuff into the infrastructure, you’re just going to be chasing your tail,” says Matt Adell, cofounder of Musical AI. “You can’t keep reacting to every new track or model — that doesn’t scale. You need infrastructure that works from training through distribution.”
Rather than detecting AI music after it spreads, some companies are building tools to tag it from the moment it’s made. Vermillio and Musical AI are developing systems to scan finished tracks for synthetic elements and automatically tag them in the metadata.
Musical AI is working on a detection system, too. The company describes its system as layered across ingestion, generation, and distribution. Rather than filtering outputs, it tracks provenance from end to end.