Fears that advanced artificial intelligence could create a biological weapon capable of wiping out humanity have grown as AI systems become more powerful. Recent research has demonstrated that AI can help design viral genomes, while companies such as Anthropic have warned that frontier models could potentially be misused to support biological weapons development. But many scientists interviewed by WIRED argue that the scenario remains far more difficult than it may appear.
The biggest obstacle is the physical world. AI can rapidly search scientific literature, combine information, and help identify laboratory protocols, but much of that information was already available through scientific publications and the internet. Creating a dangerous pathogen requires obtaining genetic material, assembling a viable genome, working with specialized laboratory equipment, testing whether the pathogen infects humans, and determining whether it can spread efficiently. Those steps require considerable expertise, infrastructure, resources, and experimentation.
Today’s laboratories also aren’t autonomous enough for an AI system to independently manufacture and release a biological weapon. Even highly automated facilities still depend heavily on humans to provide materials, operate equipment, supervise experiments, and make physical decisions. In practice, humans remain an important barrier between an AI system and the biological world.
That doesn’t mean there is no danger. Researchers are more concerned about humans using AI to make biological attacks easier than about an autonomous AI secretly engineering a pandemic by itself. AI could potentially lower some knowledge barriers, accelerate research, or help malicious actors navigate complex biological information. As laboratory automation improves, some of today’s physical limitations could also become less significant.
Experts therefore argue for layered defenses: screening orders for synthetic DNA and RNA, improving safeguards in AI models, strengthening global disease-surveillance systems, and increasing cooperation between AI companies, governments, researchers, and gene-synthesis providers. These protections could help against both AI-assisted attacks and naturally emerging diseases.
The broader point is that AI-assisted bioweapons are a legitimate biosecurity concern, but an autonomous AI creating a civilization-ending pathogen is currently considered a relatively low-probability scenario by several scientists interviewed by WIRED. At the same time, AI could significantly improve vaccine development, biological research, and responses to future outbreaks—benefits that researchers argue shouldn’t be overlooked amid fears about catastrophic AI risks.
Read the original WIRED article