Abstract

Debates about military automation and about civil-military relations turn on one question: is the place reserved for a person a working control, or a name attached to a process that runs without it? This article reads Sophocles' Antigone, in Francis Storr's translation, not only as a conflict between a ruler's edict and the unwritten laws that Antigone sets above it, but as a play about a ruler who held full authority to revoke his own order and could not use it in time. His guards cannot show whose hand buried the body and report by lot and under threat; the frightened city speaks only through his son; his seer is accused of selling prophecy; and when he yields, the living prisoner is reached last. From the play the article draws three conditions under which a stop authority is more than nominal: the human act must leave evidence that can be shown, reports must reach the one who can stop without passing through fear or dependence, and the stop must be able to reach the irreversible act first. It sets these beside Huntington's objective control, Feaver's principal-agent account and Asaro's case for meaningful human control. Two records from Meridian, a decision-record platform of the RISCON family, serve as a modern test: a committee deciding whether an AI system whose records cannot prove human approval may keep running, and a council deciding who funds and staffs the body that may stop military AI. Both are modern summaries written by AI about fictional polities. The article does not judge which option either should take.