OpenAI lays out plan for dealing with dangers of AI (Washington Post)