“OpenAI's approach to AI safety involves iteratively and gradually releasing technology into the world to allow society time to adapt and to guide research with real-world feedback.”