Sony Music's removal of more than 135,000 deepfake tracks from streaming services is being read as moderation success, but the correct reading is the opposite. The action demonstrates that a sufficiently resourced label with proprietary detection infrastructure can identify and purge synthetic imposters — once. It does not demonstrate that those songs were identified to listeners before or during playback. No streaming platform currently carries provenance labeling that tells a user whether the audio they are hearing was performed by a human or generated by a model. [Sony's target list…