AbstractExistential risk analysis has produced extensive catalogs of threats to human civilization, from asteroid impact to engineered pandemics to unaligned artificial intelligence. These threats differ in mechanism, probability, and timescale, yet they share a structural feature that the literature has noted but not foregrounded: on a one-planet civilization, every cataloged threat acts against a target with no geographic redundancy. This paper synthesizes existing observations into a unified systems framing, arguing that single-planet concentration is a common-mode vulnerability—a system-level architectural property that gives heterogeneous hazards the same terminal consequence. The paper provides an illustrative formalization using the language of reliability engineering, and extends the analysis beyond human extinction to encompass the terrestrial biological information library: approximately 3.5 billion years of evolutionary innovation, less than one percent sequenced, existing in its entirety at a single celestial address with no off-site copy. For the broad class of spatially bounded hazards, the expected loss calculation is strongly sensitive to geographic concentration, and even minimal off-site redundancy may alter loss profiles more sharply than further marginal reductions in individual hazard probabilities. The paper does not claim this holds for all risk classes; hazards with genuinely universal reach, principally unaligned AI, require separate analysis.
Ian D. Reynolds (Sat,) studied this question.