The Trump administration’s new safety review guidelines appear to exempt Chinese artificial intelligence models. OpenAI’s and Anthropic’s aren’t.