AI doomsday scenarios: Researchers weigh existential risks from bioweapons to nuclear conflict

Researchers are debating whether increasingly powerful AI systems could pose existential risks to humanity, with scenarios including bioweapons, cyberattacks, nuclear conflict, and loss of human control, though critics argue many extreme predictions remain speculative. Supporters of doomsday warnings cite examples like AI triggering nuclear war, designing harmful pathogens, or pursuing goals misaligned with human welfare, while skeptics, including cybersecurity researcher Juan Andres Guerrero-Saade, dismiss such scenarios as sci-fi that helps frontier labs recruit talent. Anthropic's Jacob Coxon, who resigned over safety concerns, says communicating specific extinction scenarios is difficult. Other discussed risks include a "runaway AI" that transforms Earth into paper clip factories, as philosopher Nick Bostrom's thought experiment illustrates, and an AI internet takeover via botnets causing chaos. RAND Corp. reports note current safeguards limit nuclear risks, but integrating AI into decision chains could change that.