Fooling the algorithm is not a robust solution. Infringement-detection algos get better hit rates all the time (fewer false negatives, although often at the pr

Fooling the algorithm is not a robust solution. Infringement-detection algos get better hit rates all the time (fewer false negatives, although often at the price of more false positives) — consider all the work that has gone into detecting musical duplicates even when resampled, pitch/time-shifted, rerecorded, etc. The robust solution is to treat files which are redistributed and therefore trigger copyright provisions completely differently from files which are not redistributed. Furthermore, for the purposes of punitive action, sharing a copyrighted file with a limited whitelist of other users might reasonably be treated as a less of a screwup than making a file accessible to the entire internet. And therefore it should not be frictionless to share a file with the internet.