The world is fragmenting. So will superintelligence.
Most AI safety work assumes one superintelligent system, a single AI that either saves or destroys us. That is not where the trajectory points.
The US and China are building advanced AI on parallel tracks. Multiple frontier labs are approaching comparable capability levels simultaneously. Open-source models are spreading globally. And the geopolitical order that assumed stable cooperation is breaking apart in real time.
The most likely outcome when AI becomes superintelligent is not one AI ruling the world. It is Fragmented Superintelligence (FSI), multiple superintelligent systems, backed by competing states and corporations, locked in sustained strategic competition.
Not one AI to rule as the Singularity, but many AIs, competing with each other.
And the danger is not any single system. It is the competition itself.
Every lab already faces the question: spend another year on safety, or deploy before your competitor captures the advantage? Every government asks: constrain our AI programme, or fall behind a rival with no constraints? These pressures do not disappear at superintelligence level. They become existential.
But what we need to seriously consider now is this. Does this competition produce real human choice between accountable systems? Or competing AI sovereigns carving up the world, where your only "freedom" is choosing which one controls your life?
Without the literacy to evaluate these systems, the institutions to hold them accountable, and the cross-cultural capacity to cooperate across AI-governed boundaries, it is not freedom. It is feudalism with better marketing.
Almost nobody is building for that. We are.