The Controllability Trap: A Governance Framework for Military AI Agents
The article discusses a new governance framework for military AI agents called the Agentic Military AI Governance Framework (AMAGF). It identifies six governance failures that can occur with agentic AI systems and proposes a continuous model for measuring and managing control quality. The framework aims to enhance human oversight in military operations through preventive, detective, and corrective governance strategies.
- ▪The framework addresses distinct control failures not covered by existing safety measures.
- ▪It introduces a Control Quality Score (CQS) to quantify human control in real-time.
- ▪The governance model emphasizes continuous measurement and management of control quality throughout the operational lifecycle.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Computers and Society arXiv:2603.03515 (cs) [Submitted on 3 Mar 2026] Title:The Controllability Trap: A Governance Framework for Military AI Agents Authors:Subramanyam Sahoo View a PDF of the paper titled The Controllability Trap: A Governance Framework for Military AI Agents, by Subramanyam Sahoo View PDF HTML (experimental) Abstract:Agentic AI systems - capable of goal interpretation, world modeling, planning, tool use, long-horizon operation, and autonomous coordination - introduce distinct control failures not addressed by existing safety frameworks. We identify six agentic governance failures tied to these capabilities and show how they erode meaningful human control in military settings.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv.org.
Discussion
0 commentsMore from arXiv.org
