The existential risks posed by powerful AI systems demand serious safeguards. But our nation should not treat every use of AI as equally dangerous.