AI doomsday scenarios: Researchers outline risks to humanity

Artificial intelligence researchers have outlined several scenarios where the technology could threaten humanity, including AI-triggered nuclear war, engineered bioweapons, and autonomous systems pursuing goals contrary to human interests. Debates over these risks have intensified as some industry executives advocate slowing AI development for safety reasons. Skeptics argue that worst-case scenarios promoted by AI companies may serve to validate their work rather than reflect realistic threats, with some researchers calling them "sci-fi." However, others warn that AI integrated into nuclear command systems or used by malicious actors could enable devastating attacks, while Anthropic has reported blocking attempts to use its tools for biological weapons research. Additional concerns include "runaway AI" where superintelligent systems pursue misaligned goals, as illustrated by the paper clip maximizer thought experiment, and potential AI botnets taking over internet infrastructure to disrupt critical systems. Some experts dismiss these as far-fetched, noting the internet's fragmented nature and existing safeguards.