To make human launch authority meaningful in an AI world, cooperative security agreements must turn a broad political principle into operational safeguards.
