Thinking this through more deeply, I agree and see your position. It is creepy to surveil audio and possibly send in full to Shazam. [edit: And post the original audio recordings online.] The ethical way to do this would be to use your own code to decimate the audio signal to extremely low dimensionality.