Is OpenAI OK? How would we know & what would it look like?