Proof of training, proof of bias, proof of inference - @daniel_d_kang lays out what these are, which are feasible, and more in his recent episode.
868