Dual-use capability controls and catastrophic risk limits

Dual-use AI capabilities—those useful for both beneficial and harmful purposes—drive both export-control style restrictions (like blocking unauthorized model distillation to protect sovereign AI advantages) and technical safeguards (like biosecurity software barriers) meant to keep capabilities below catastrophic risk thresholds. These terms collectively show the tension between nations wanting to control AI capability spread and the technical guardrails meant to prevent misuse of that capability.