Published by RAND Corporation in 2025, this scenario-based analysis benchmarks AI's extinction-level risk against other technologies historically capable of causing human extinction, including nuclear weapons, engineered pathogens, and geoengineering. It concludes that AI causing human extinction would be extremely difficult under current and near-term capabilities, though the report stops short of ruling the possibility out entirely.
The report's methodical, comparative framing has made it a frequently cited reference point in policy debates that otherwise tend toward more speculative extremes on either side of the AI-risk discussion.