Comment by Rogerio Piovesan

This article therefore recommends that the United Nations General Assembly formally open a pathway toward an International Artificial Intelligence Agency. Governed by its own statute, the agency would have a dual mandate for promotion and verification, with enhanced scrutiny limited to frontier AI activities exceeding internationally agreed thresholds for risk, autonomy, reach or cross-border impact. [...] Even so, frontier models depend on concentrated compute, advanced semiconductors and large blocks of firm power – costly, licensed and metered assets that create potential observation points.