Scientists Used AI to Build Viruses From Scratch and 16 of Them Actually Work

A First in Science: Viruses Written by Machine

Artificial intelligence has been used to design brand new viruses that are fully functional and can replicate in the laboratory — and it marks the first time whole genomes have been successfully designed by AI. The study, published August 6 in the journal Science, is being called a landmark moment in biology, one that could reshape how medicine fights its most stubborn enemies.

Led by Brian Hie, assistant professor at Stanford University and an innovation investigator at Arc Institute, the researchers used two genome language models — Evo 1 and Evo 2 — to generate complete phage genomes with realistic genetic architectures and specificity for a bacterial host, Escherichia coli. The two systems learned the chemical alphabet of DNA from 2.7 million genomes, then generated whole viral blueprints letter by letter — blueprints no living cell had ever carried.

Hie described the process simply: "In this case, we wanted the model to generate the entire genome end-to-end in a single left-to-right pass. We didn't add anything." The result was thousands of novel viral designs, none of which exist anywhere in nature.

Sixteen Viruses That Can Kill Drug-Resistant Bacteria

Scientists from Stanford and the Arc Institute used the AI to generate thousands of genome combinations within a specific framework — one that could be hosted by an E. coli bacteria cell. Of the thousands of genomes generated, scientists built and tested about 300 in the lab and found that 16 were viable viruses. Some of the designed phages even outperformed the natural E. coli-infecting phage in killing bacteria in direct laboratory competition.

The implications for antibiotic resistance are significant. As Hie explained, "If the bacteria gain resistance to a single phage, it's game over for the medication. But if you have multiple genetically distinct phages in a mixture, it would be harder for the bacteria to develop resistance to the entire cocktail." The team tested a combination of the 16 phages against E. coli that had already become resistant to the natural phage — and showed the cocktail rapidly overcame that resistance.

The study represents a significant advance in generative genomics, a field seeking to use computational models to create functional biological systems rather than modifying one gene or a small genetic circuit at a time. Drug-resistant infections are projected to kill roughly 40 million people worldwide by 2050, making this line of research urgent. The AI's exploration of sequence space provides raw material for rapid adaptation to resistance mechanisms, potentially transforming phage therapy from a trial-and-error process into a systematic approach for staying ahead of bacterial evolution.

A Powerful Tool That Demands Careful Oversight

The breakthrough has been labeled a "very significant turning point" in science that could unlock a new era for treating disease — but experts have also warned that AI-designed viruses raise "urgent" safety and security concerns. In an accompanying Perspective published in Science, Thomas Inglesby and Moritz Hanke of Johns Hopkins University said the advance raised biosecurity concerns, calling for legally required screening of synthetic genetic-material orders for potentially dangerous sequences instead of relying on voluntary safeguards.

The biosecurity warning published alongside the work names two fixes on the table: a legal duty for synthetic-DNA providers to screen every order and customer, and new detection tools tuned to catch AI-written genomes that match nothing in nature. Neither exists in required, deployed form today. Since the Evo 2 model is fully open — including its training code, inference code, model parameters, and training data — a sufficiently motivated person could, in principle, adapt these models toward other ends.

The Arc Institute noted that this work was conducted with comprehensive safety protocols that exceeded standard requirements. All experiments took place in dedicated biosafety cabinets with specialized disposal procedures, and equipment never left the containment area. Only non-pathogenic bacterial hosts were used, ensuring both the original and AI-designed bacteriophages posed no risk to human health. As one commentary put it, "The question is no longer whether generative viral genome design will exist. It is whether society can build oversight that allows its benefits to unfold while preventing it from enabling serious harm." That question may define the next chapter of this technology just as much as the science itself.