When Everything Becomes a Weapon

Military technology offers the starkest illustration of what happens when AI meets institutional power, because warfare strips away the ambiguity that surrounds civilian applications. In commerce, the consequences of a bad AI decision might be a misdirected advertisement or an unfair lending decision. In war, the consequence is death.

Weapons have always incorporated forms of automation. Guided missiles follow predetermined trajectories. Radar systems track objects and calculate intercepts. Defensive systems like the Phalanx close-in weapon can automatically engage incoming threats faster than any human operator could react. Automation in warfare is not new.

What is new is the prospect of moving automation further up the decision chain, from the mechanics of engagement to the logic of targeting itself.

Instead of merely answering "Where is the target?" (a question of tracking and navigation) AI systems are increasingly positioned to help answer a different set of questions: "Which object is the target? Is it hostile? How important is it relative to other potential targets? What response is proportionate?" These are judgment questions, not engineering questions. They involve interpretation, context, and the weighing of competing considerations. And they are being progressively delegated to machines.

A 2025 analysis by the Stockholm International Peace Research Institute examined this problem directly, comparing autonomous weapon systems with AI-enabled military decision-support tools and warning about the risks of unintended harm as AI becomes integrated into successive stages of military targeting.

The distinction between a weapon and a decision-support system is becoming dangerously blurred. Consider a plausible scenario: a commander receives a list of ten potential targets. An AI system has identified, ranked, and prioritized them. The commander approves the first recommendation. The system suggests the appropriate weapon. The commander approves. The system calculates the delivery route. The commander authorizes launch.

At every stage, a human being has technically remained in the loop. A human hand has pressed every button. But how much of the underlying decision was genuinely human? The targets were machine-selected. The prioritization was machine-generated. The weapon choice was machine-recommended. The human's role, in practice, was to ratify a chain of conclusions produced by a system operating at speeds and scales that no human mind could independently verify.

This is the central governance problem of AI-enabled warfare: human oversight can become a formality rather than a meaningful check. The loop can be preserved in form while being emptied of substance.

The United States is already grappling with these questions at the policy level. The Congressional Research Service describes lethal autonomous weapon systems as those capable of independently identifying a target and employing an onboard weapon without manual human control. In 2026, Congress saw legislative proposals specifically addressing the use of AI in decisions involving lethal force, domestic surveillance, and nuclear weapons.

The debate is no longer theoretical. The systems are being designed, funded, tested, and in some cases deployed. The infrastructure of a new kind of warfare is under construction, and the rules governing its use remain incomplete.