Description
Security's contemporary condition is legible in the collapse of the dichotomies structuring international politics, predominantly in their Western articulations: war and peace, civilian and military, inside and outside. Both Walt's classical definition and the Copenhagen School's widened notion assume civilian and military can, in principle, be divided. Dual-use technologies and operations conducted short of war dissolve that assumption, rendering securitization indistinguishable from desecuritization and leaving the civilian sphere undecidably Arendtian or, as Foucault argued, already the continuation of war by other means. This collapse extends to the competition/cooperation distinction structuring existing international frameworks.
Large language models radicalise this condition. Embedded within security fields, they fuse framing — Butler's apparatus determining what becomes recognisable as threat — and practice — Bourdieu's and Bigo's routinised work through which professionals produce insecurity — into a single computational pass. Their apparent predictive superiority breeds structural deference, producing "moral crumple zones" where humans nominally retain oversight while no longer exercising judgement.
International relations theory's inherited vocabulary, built for separable domains, cannot adequately grasp a condition of security that technology has rendered total. The paper proposes to treat this impossibility as the condition for rethinking conflict and cooperation in the age of AI.