Researchers have reached a major milestone in synthetic biology by using AI to design complete viral genomes, convert those digital sequences into DNA, and produce functional viruses in the lab.
The work was led by teams from Stanford University and the Arc Institute, who used the Evo 1 and Evo 2 models to explore DNA patterns and generate candidate genomes. The focus was a small bacteriophage related to ΦX174, a virus that infects E. coli and poses no risk to humans.
How the experiment worked
Rather than asking a general-purpose model to "invent" a virus, the researchers trained the system on thousands of related Microviridae genomes. After computational filtering, they selected 285 designs for lab testing. Most did not work, but 16 produced viruses capable of infecting bacteria and making copies of themselves.
Some of those AI-designed phages even outperformed the natural ΦX174 reference in laboratory tests. In mixed samples, they also helped overcome bacterial resistance, pointing to a possible future role in phage therapy.
That matters because phages may offer a targeted alternative for treating bacterial infections that no longer respond well to antibiotics. AI could expand the search for useful viral candidates and help scientists build tailored phage combinations more efficiently.
A powerful tool with broad implications
The study also highlights how quickly biological AI is advancing. Evo 2 is publicly available, which makes the technology more accessible to researchers around the world. At the same time, it shows how generative models can reduce the distance between a digital sequence and a living system.
The researchers emphasize that this is still a laboratory proof of concept, not a ready-made medical solution. Even so, the result demonstrates that AI is moving from predicting biology to actively helping write it.
Published in Science, the study suggests that future breakthroughs in medicine, biotechnology, and genetic design may arrive faster as AI becomes more capable and more precise.
