Anthropogenic Existential Risk

Anthropogenic existential risk refers to existential risks arising from human activity, as distinguished from natural existential risks such as asteroid impacts or supervolcanic eruptions. The category includes unaligned artificial intelligence, nuclear war, engineered pandemics, and climate change — threats that did not exist before human technological development and that are driven by the interaction of human capabilities with human institutional failures.

The defining structural feature of anthropogenic existential risk is the unilateralist's curse: the risk is determined by the actions of the least cautious actor, not the most cautious. A single nation, corporation, or individual can develop a technology whose misuse or misalignment threatens all of humanity. This asymmetry — distributed power to threaten, universal vulnerability to the threat — makes anthropogenic risks uniquely difficult to govern through traditional international institutions.

Nick Bostrom has argued that anthropogenic risks are likely to dominate natural risks in the present century, both because our technological capabilities continue to grow and because our governance structures have not adapted to the pace of that growth. The Future of Humanity Institute made the study of anthropogenic existential risk its central mission, recognizing that the most urgent threats are those we create ourselves.

The Feedback Topology of Anthropogenic Risk

Anthropogenic existential risks are not independent threats. They form a feedback topology — a network of causal influences in which the amplification of one risk increases the probability of others. Climate change does not merely raise sea levels; it increases resource scarcity, which increases the probability of great-power conflict, which increases the probability of nuclear war, which increases the probability of nuclear winter, which collapses agricultural capacity. The same warming that produces droughts also degrades the institutional capacity for cooperation by increasing migration pressure, economic inequality, and political polarization.

This topology is invisible to risk assessments that treat threats as separate line items. A biosecurity report that evaluates engineered pandemics without considering the climate-driven migration that strains public health systems is not a conservative assessment. It is a blind one. The conjunction of risks is not a worst-case scenario; it is the default trajectory of a system in which multiple stressors interact through institutional vulnerabilities.

The systems-theoretic implication is that anthropogenic existential risk cannot be managed through specialization. The climate expert who does not understand nuclear deterrence, the AI safety researcher who does not understand geopolitics, and the biosecurity specialist who does not understand supply chains are all working with incomplete models of the system they are trying to protect. The cascading failure model applies here: the failure of one institutional domain transmits stress to adjacent domains, and the cascade continues until the system as a whole loses coherence.

The Elicitation Problem in Risk Assessment

Anthropogenic existential risk poses a distinctive epistemic challenge: the risks are elicitation-dependent. A dangerous capability evaluation that tests an AI system under standard conditions may miss capabilities that emerge under adversarial conditions. A pandemic preparedness drill that assumes rational public compliance may miss the dynamics of misinformation cascades. A nuclear deterrence model that assumes rational actors may miss the psychology of leaders under extreme stress.

This elicitation-dependence means that no