As generative AI capabilities continue to advance, concerns over an "AI-driven bioweapon catastrophe" have visibly intensified in recent weeks. Heads of several AI companies had earlier called for tighter oversight of synthetic DNA manufacturing, and researchers at Stanford University and the Arc Institute demonstrated that AI can design novel viral genomes — developments that have further fueled discussion of "AI-engineered lethal pathogens."
However, a number of scientists argue that AI's practical threat in the bioweapon realm remains limited for now. Researchers note that AI functions more as a tool for accelerating information retrieval and synthesis, and has not fundamentally altered the key bottlenecks in bioweapon development. The real hurdles still lie in the laboratory: assembling a complete genome from scratch, verifying whether it can infect humans and transmit between people, and acquiring the raw materials, equipment and specialized expertise required.
AI's threat to bioweapon development remains limited
From a bench-level perspective, AI today also struggles to carry out such workflows autonomously. Executives at biotechnology firm Ginkgo Bioworks note that even if AI can participate in experimental procedures, it still requires humans to supply equipment, reagents and hands-on support, and human review itself serves as a significant check. Some experts also argue that even if more powerful general-purpose AI emerges in the future, human researchers will still be able to respond through vaccine development and public-health defense systems.
Other voices caution that in the near term, what warrants greater vigilance is not AI independently launching a biological attack, but rather its exploitation by bad actors. Relevant experts note that although fully autonomous laboratories are not yet widespread, AI may lower certain thresholds by aiding information integration, task decomposition and execution assistance, so the possibility of its use in bioterrorism should not be dismissed. Other researchers point out that bioweapons themselves are not an efficient mode of attack; the difficulties lie in precise, large-scale manufacturing and transmission control, and real-world operational complexity often exceeds that of conventional attack methods.