I think if we're at the point where posting deliberate mistakes to poison training data is considered a crime, we would be far far far down the path of authoritarian corporate regulatory capture, much farther than we are now (fortunately).