The Convention

We are entering an arms race with no rulebook. Unlike nuclear weapons, which required rare materials and immense state capacity to develop, military AI is being built largely with commercial technology — the same machine learning techniques powering consumer apps are being repurposed for targeting systems and battlefield analysis. That accessibility should worry us more than it currently does.

International institutions have historically been slow to regulate emerging military technology, and AI is no exception. The UN’s discussions on “lethal autonomous weapons systems” have continued for over a decade without producing a binding treaty. Meanwhile, the technology has moved from theoretical concern to operational reality.

This is not an argument against AI in defense contexts — that ship has sailed, and pretending otherwise is not a serious policy position. It’s an argument for urgency. The states currently leading in military AI development have the most to gain from delaying regulation, and the most leverage to shape what regulation eventually looks like. Waiting for a catastrophic failure before acting would be a failure of imagination the world can’t afford twice.