The copyright cases advancing through U.S. courts — including the claims against Stability AI now proceeding to discovery — are structured around training data: who owned what was fed into the model. That framing leaves the authorship question untouched by design. A court finding that Stability AI infringed in training its model says nothing about whether a person who types a prompt into that model has created something. Practitioners who frame prompting as a form of commissioning — "you could always pay someone to do what you cannot" —…