Copyright Law Has a Test for AI Music. It May Be Asking the Wrong Question.

Matthew Sag did not argue that AI music companies are innocent — he argued that copyright may simply not be equipped to find them guilty. His formulation at the Berkeley symposium, captured and shared on Bluesky ^[1], is deceptively simple: if a model's outputs are not substantially similar to the works it trained on, copyright's primary enforcement mechanism cannot reach the practice. This is not a loophole argument. It is a structural observation about what copyright was designed to do and what it was not.