As AI agents demonstrate increasingly autonomous behavior, Gabriel Accascina proposes a UN-led evaluation system modeled on WHO prequalification, using independent testing and government procurement incentives to strengthen international oversight of frontier AI.