The Commander-in-Chief Contradiction: On the Incoherence of Scaling-First AI Security Policy, and the Deterministic Alternative

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

The current administration has adopted a position on artificial intelligence that is internally contradictory on its own terms. It declares AI a national security imperative, yet endorses systems that cannot be certified for any security-critical deployment. It dismisses AI safety warnings as conspiracy, yet has signed security-testing instruments and convened meetings on the challenges of scaling. It proposes kill switches for a system it simultaneously claims requires no restrictions. This paper argues three things:1. The contradiction is not confusion. It is a position engineered to be unfalsifiable — tocapture the political and commercial benefits of a security framing without accepting theaccountability that security implies.2. The energy wall is not a climate problem that happens to touch AI. It is a security problem, because concentrated, grid-dependent, cloud-resident inference is structurally indefensible.3. A bounded alternative resolves both. ST-T1024 is deterministic by construction, certifiable under standards already in force, air-gappable, and reduces per-query energy by a de-rated 97.8%. It does not compete with frontier AI. It occupies the category the security state actually requires and cannot currently procure.