AI tools can now generate full songs, raising new copyright fights

In short: AI can now create complete songs from a short text prompt, and the music world is scrambling to figure out the rules and what listeners will accept.

What's going on

AI music tools can generate an entire track, including lyrics, singing, and the finished sound, from a few words. Think of it like ordering a custom song the way you order a sandwich, you describe what you want and it arrives. Systems mentioned in recent reporting include Google DeepMind’s Lyria, and popular tools like Suno and Udio.

These tools are improving fast, but they still have gaps. Long songs can lose structure, vocals can sound less natural than a strong human performance, and the music can feel safer and more predictable. Researchers are also starting to measure whether AI music matches what people actually like, not just whether it sounds polished.

The biggest fights are about rights and ownership. In the US, the Copyright Office says it will not register a work made entirely by a machine with no human creative input. That can make it hard for someone to stop others from copying an AI-made song. At the same time, record labels are reported to be negotiating licensing and revenue-sharing deals with some AI music companies, partly to address lawsuits and questions about training data (the songs used to teach the system, like study materials for a student).

What to watch

Expect more debates about voice cloning, when a tool copies a singer’s sound so closely that it can mislead listeners. Also watch for labeling efforts such as audio watermarking, which is like an invisible stamp inside a file that can help platforms identify AI-made tracks. Finally, keep an eye on where AI music shows up most, background music for ads and games may change faster than fan-driven pop.

Source: NYTimes