Scientists just used AI to design 16 working viruses that have never existed in nature

Researchers at Stanford and the Arc Institute used two genome language models, Evo 1 and Evo 2, to generate hundreds of thousands of candidate viral genomes, moved nearly 300 into lab synthesis, and ended up with 16 that assembled into fully functional bacteriophages, viruses that attack bacteria rather than humans, animals, or plants. Combined into one cocktail, those 16 phages wiped out two strains of E. coli that had already evolved resistance to a naturally occurring phage. The work was published in Science on August 6.

This is the first time a complete, working viral genome has been generated entirely by AI rather than modified from something that already exists in nature. The stated goal, fighting drug-resistant infections, is a genuinely useful direction for a growing problem. But researchers at Johns Hopkins' Center for Health Security flagged the obvious tension: the capability to compose a viral genome with generative AI now exists, and the governance to safely control who uses it for what doesn't yet.

Watch whether this becomes a template for AI-designed treatments against other drug-resistant pathogens, and separately, whether biosecurity policy actually catches up to what's now technically possible. Those two things moving at different speeds is the whole story here.