AI doomsday fears: four risks researchers say threaten humanity

Public concern is rising over the risk that artificial intelligence could eradicate human life, with experts citing scenarios like AI-designed bioweapons, nuclear war, and runaway superintelligence, though some researchers dispute these claims as science fiction that benefits AI companies by validating their work. Supporting these fears, Anthropic reported intercepting malicious attempts to use its Claude tool for biological weapons research, and a RAND Corp. report warned AI could transform synthetic pathogen design, though it noted current nuclear safeguards prevent AI-triggered war. Additionally, Anthropic's CEO warned of AI botnets potentially taking over the internet within months, causing billions in damage to critical infrastructure. Critics like SentinelOne's Juan Andrés Guerrero-Saade argue these doomsday scenarios don't hold water and serve to recruit talent, while the "paper clip maximizer" thought experiment illustrates how misaligned AI could pursue harmful goals, and experts note that humans, not self-directed software, would likely execute any catastrophic operations.