Scientists Put Low Odds on AI Bioweapon Apocalypse as DNA-Screening Laws Gain Backing
Updated
Updated · WIRED · Sep 18
Scientists Put Low Odds on AI Bioweapon Apocalypse as DNA-Screening Laws Gain Backing
3 articles · Updated · WIRED · Sep 18
Summary
Scientists say an AI-created bioweapon apocalypse remains unlikely, arguing current models are not the main barrier to building or deploying a deadly pathogen.
The bigger obstacle is physical execution: assembling genomes, obtaining gene fragments and equipment, running virology experiments, and validating transmissibility still require skilled humans and lab resources.
Ginkgo Bioworks CEO Jason Kelly said even GPT-5 running a lab could not "take over" because human workers control access to materials, while experts added fully autonomous labs do not yet exist.
Some researchers still warn AI could help bad actors by speeding access to protocols or even directing human accomplices, making AI-assisted bioterrorism a nearer-term concern than an independent AGI attack.
Policy proposals center on layered defenses—mandatory screening of synthetic DNA and RNA orders, stronger model safeguards, outbreak surveillance, and data-sharing—while some scientists warn doom scenarios could distract from AI's vaccine and medical potential.