tylerjohnston comments on On the Meta and DeepMind Safety Frameworks