Stanford researchers create artificial viruses to target bacteria using AI

Researchers from Stanford University and the Arc Institute have used the AI model Evo to design complete viral genomes, producing 16 functional viruses that do not exist in nature. The study, now peer-reviewed in Science, involved Evo proposing 700,000 possible genomes, of which the team pursued the most promising, synthesizing 285 sequences as DNA and introducing them into bacteria. Of these, 16 produced replicating viruses. Evo was first trained on about nine trillion nucleotides from millions of organisms, then on the 11 genes of the phage Phi X-174 and about 15,000 relatives. The created viruses were as robust as natural ones, with some replicating faster than Phi X-174. However, critics note the viruses resemble natural species, and the team deliberately excluded human-infecting viruses from Evo's training to avoid generating dangerous pathogens. The work highlights a regulatory gap, as current US NIH guidelines do not cover pure computer-based virus genome design.