Stanford team creates AI-designed viruses not found in nature

Stanford University and Arc Institute researchers have synthesised self-replicating viruses with genomes generated by AI, marking the first reported creation of brand-new viruses not found in nature. The team used genome language models Evo 1 and Evo 2, which predict genetic code after training on genomes from viruses, bacteria, plants and animals, to design bacteriophages able to infect E. coli.

Using ΦX174 as a template, the models generated about 700,000 potential designs, and researchers selected 285 for synthesis and testing. After the DNA was inserted into E. coli, 16 Petri dishes showed clear spots as viable phages attacked and replicated inside the bacteria. Some performed better against E. coli than the original ΦX174 bacteriophage.

The work points to possible uses in phage therapy for antibiotic-resistant bacteria and in custom enzyme design for genetic disorders, but it also raises biosafety and biosecurity concerns. Researchers excluded viruses that infect humans and animals from training data, worked in a secure lab, and chose ΦX174 because it only attacks E. coli. Johns Hopkins Center for Health Security researchers warned that governance has not kept pace, while Evo 2 has been made freely available to the public.