Anthropic pulled two new AI models offline after the Trump administration said they were too dangerous for foreigners to use. Rival OpenAI agreed to let the administration vet users of its newest model.