AI Bioweapon Fears Overblown, Experts Say—Real Bottleneck Is Biology, Not Algorithms

Fears that rogue artificial intelligence could unleash designer pathogens and wipe out humanity have gripped policy circles from Silicon Valley to Washington. Yet many scientists argue this scenario grossly underestimates the practical obstacles between having information and actually building a deadly virus.

The Real Barrier Isn't Data Access#

Researchers have long debated how openly to share biological knowledge—a dilemma that predates AI. David Bellamy, a scientist at the Institute of Foundation Models, points out that the internet, academic journals, and translation tools already made lab protocols and techniques widely available before large language models arrived.

AI systems can accelerate the speed at which researchers find and synthesize this information, but that acceleration alone doesn't overcome the fundamental challenge: assembling a functional pathogen. "AI is a tool that can help both good actors and threat actors access information faster," Bellamy explains, "but those capabilities are not the bottleneck in bioweapon production."

The Physical Bottleneck: Where Theory Meets Reality#

The true constraint lies in the wet lab. Constructing a virus from scratch requires gene fragments, specialized equipment, and technical expertise. A would-be bioweapon creator must synthesize a complete genome, verify it can infect humans and trigger disease, and ensure person-to-person transmissibility. Even with robotic lab assistants handling routine tasks, humans remain essential to troubleshoot failed experiments and make judgment calls.

Jason Kelly, CEO of biotech firm Ginkgo Bioworks—which recently ran a project where OpenAI's GPT-5 system controlled laboratory equipment—is skeptical an AGI could independently complete this chain. "An AI couldn't commandeer the lab," he says, partly because human workers could simply refuse to provide substances and equipment a rogue system requests. Scaling autonomous bioweapon creation would require "dramatically more robots all over the place" to bypass that human control layer.

The Vaccine Counter-Threat#

Even in a worst-case scenario where an AI-designed supervirus escaped containment, immunologist Derya Unutmaz argues the threat would be containable: researchers could deploy AI systems to rapidly design and manufacture countermeasures. The speed advantage flows both directions—defense as well as offense.

Where Genuine Concern Remains