The Sovereign AI Myth

Abstract

One morning last winter, the International Criminal Court’s chief prosecutor discovered his official email had stopped working. The reason? Overnight, Microsoft shuttered his account to comply with new Trump Administration sanctions. The incident showed the world how dependence on foreign technology could be weaponized on a whim. States now believe that for vital technologies like artificial intelligence, they can no longer depend on foreign suppliers; they must build them themselves.

This Article uses the growing clamor for “sovereign AI” as a window onto a deeper transformation. Recent scholarship has documented how states seek immunity from external law in the digital realm. This Article shows how that pursuit leads inexorably to another: the quest for immunity from reliance on external sources of technological supply. Since every external dependency is a channel through which foreign leverage can reach in, every such channel must be closed. The endpoint is autarky, and its costs are ruinous.

Worse, sovereign AI discourse forgets how interdependence can foster accountability, even in war zones. Months after the ICC incident, Microsoft terminated AI services an Israeli intelligence unit had been using for mass surveillance in Gaza. Had Israel developed its own “sovereign AI,” nothing could have been done.

This Article argues it is a mistake to sever the channels of interdependence through which both coercion and accountability can flow. The better path is to rebuild international law to prevent the former while permitting the latter, and to bind the powerful to those rules.

Keywords: artificial intelligence, international law, human rights, sovereignty, supply chains, digital sovereignty, international economic law, globalization