What are good safety standards for open source AIs from China? — LessWrong