Autonomy at War: Proportionality and the AI Race in Future Warfare

As nations accelerate the development of autonomous weapons systems infused with artificial intelligence (AI), the balance between technological advantage and moral restraint grows increasingly fragile. Programs such as the U.S. Department of Defense’s Replicator Initiative and the rise of private defense innovators like Shield AI and Anduril illustrate an urgent race to deploy systems capable of operating in GPS-denied and communication-degraded environments. While these advances promise speed, precision, and deterrence, they also challenge the International Humanitarian Law (IHL) principle of proportionality, the requirement that civilian harm not be excessive in relation to anticipated military gain. The opacity of machine-learning decisions, the risk of hallucination or misidentification, and compressed testing timelines that reduce human oversight erode confidence that lethal decisions remain ethically constrained. The recent public dispute between Anthropic and the U.S. Department of Defense over explicit guardrails on fully autonomous weapons further underscores that these concerns are no longer theoretical, but active points of contention in defense AI procurement and governance. This article argues that maintaining human judgment at the point of lethal decision is indispensable until auditable and verifiable AI systems exist. It concludes with policy recommendations for rigorous verification, transparent testing, and the adoption of multilateral governance frameworks.