AI: COULD ANYONE INSPECT IT AND STOP IT? On Daniel Suarez's Warning, the Authoritarian Condition, and the Architecture That Passes the Test

Science fiction writer Daniel Suarez warns of AI-powered authoritarianism:"Do you really want your worst idea of a leader, whether it's Trump or somebody else, tohave a robot army and complete surveillance 24/7, driven by AIs that watch everything allthe time?" The warning is precise. It is also incomplete — because it names the danger without namingthe test. This paper states the test. It is the only test that matters when a capability is concentrated:could anyone inspect it after the fact and stop it? The paper applies the test to the current architecture. It shows that a stochastic system fails it at every level — no specification to inspect against, no trace required, no signer to stop. It then applies the test to the deterministic alternative and shows that it passes by construction. The paper does not argue that authoritarianism is inevitable. It argues that the architecture determines whether authoritarianism is even inspectable — and that the current architecture makes it invisible.