The most revealing detail in The Verge's investigation into Suno's copyright nightmare is not the obvious one. It is not that the platform generates material that sounds like copyrighted work. It is that Suno's own policy creates an impossible standard: no copyrighted input, but upload your own tracks to remix. The self-referential contradiction is the story. You are both protected and the risk, simultaneously, by design.
AI Music, Benchmark Plateaus, and the Evaluation Problem
This is not just a legal question. It is an evaluation architecture question. A 2026 arXiv paper on XpertBench documents how LLMs are hitting plateaued performance on conventional benchmarks because the benchmarks were not designed for the complexity of real expert tasks. Music copyright law is exactly that kind of complex expert task. The legal standard for infringement involves substantial similarity, intent, market substitution, and cultural context. No current AI system, including the ones advising on Suno's policy, can reliably evaluate all four simultaneously. The policy is written for a world where evaluation is simple. The world is not.
The New Deal for Sounds: Labor, Capital, and the Sample Economy
The ARTnews piece on the New Deal's vision of artists as workers keeps returning as a ghost in this conversation. Sample culture, remix culture, and now AI music generation are all downstream of an unresolved question about who owns the cultural raw material. The WPA paid artists to produce it. The streaming economy monetized it. AI platforms are now consuming it and calling it transformation. TurboFund's live investor signals show AI music and audio startups as one of the most active seed categories in 2026. The capital is moving fast. The legal infrastructure is moving at WPA speed.