RedMan comments on The map of “Levels of defence” in AI safety