Stanford and Arc scientists used AI to design bacteria-killing viruses

A team from Stanford University and the Arc Institute used an AI model to design complete viral genomes from scratch, then built 16 functional viruses in the lab that do not exist in nature. The work, previously available as a preprint, has now been peer-reviewed and published in Science, according to The Decoder. The report describes the project as an early step toward AI-designed life forms, while noting that viruses are not technically alive.

The model, called Evo, proposed 700,000 possible genomes, and researchers pursued only the most promising candidates. The team had 285 sequences chemically synthesized as DNA and inserted them into bacteria; 16 produced viruses capable of replicating. The original preprint had mentioned 302 synthesized genomes, while the updated report adds detail on the hit rate and peer-reviewed publication.

The report points to possible applications in phage therapy, where viruses are explored as tools against multidrug-resistant bacterial infections. It also notes that viruses are important in gene therapy because they can deliver genes into human cells. Scientists cited in the article described the work as an important milestone, but also stressed that the viruses were very similar to natural species and relied on the same biology.

The article highlights a biosafety gap: current U.S. policy on high-risk life sciences research bans experiments that make pathogens more dangerous, but purely computational viral DNA design is not covered unless it involves an entity of concern. The Stanford and Arc team avoided training Evo on viruses that infect humans, as well as related pathogens from animals, plants, or fungi. The larger concern is how oversight should handle AI-generated biological designs that may not fit existing categories.