The safety people aren't that bright. They are just trying to write tons of requirements that the actual ML model writers follow to make sure race and violence is accounted for somehow. These are not AI experts. They are mostly lay people.