Decision sovereignty and defense AI in the Global South: the Brazilian case

Abstract

This article identifies, from the Brazilian case, conditions for preserving decision sovereignty in the development of a Defense artificial intelligence (AI) structure in Global South contexts marked by technological vulnerabilities and communication constraints in contested networks. It adopts a predominantly qualitative approach, supported by descriptive quantitative indicators, and uses Brazil as an embedded single-case study, with the United States as a contextual contrast. The corpus combines defense legislation and policy, military doctrine, multilateral documents, academic literature, technology-readiness indicators, cloud-infrastructure data, and records of firms accredited by the Brazilian Ministry of Defense. The analysis treats external technological dependence and algorithmic opacity as distinct problems. In Brazil, external exposure is clearest in computing infrastructure, while capabilities related to AI models remain less clearly determined by public sources. Documented capabilities are more visible in integration, communications, information security, and command and control. The analysis of military decision-making shows that opacity, cognitive load, trust calibration, and degraded communications condition the assessment of AI-generated outputs. The findings are organized around control and authority, verification, explainability, and continuity. The article concludes that decision sovereignty does not require complete domestic production of the technological chain, provided that the state preserves sufficient capabilities to control, assess, and sustain the functions that condition the military use of AI.