AI's threat to human beings lies not in its continuous upgrading in intelligence and the eventual evolution of "human?like" consciousness， but in human's intentions it bears. Advances in neural network technology have showed that humans are replicating their intentions onto AI vehicles. The current wave of AI militarization embodies humanity's negative intentions， which will upend the military balance among great powers， and more importantly， will lead to the twinning of enmity among nations into AI systems， thus posing a serious threat to mankind as a whole. To response， the international community should take a common security as guiding principle， take a community with a shared future as value direction， build a knowledge community based on AI， judge AI's technological inflection point from scientific level， and alleviate the security dilemma through international arms control mechanism and inter?state agreements， in an effort to accurately manage and control the security challenges of AI militarization in parallel multi?tracks.