AI wrote 16 working viruses from scratch — and the study just cleared peer review in Science
Stanford and Arc Institute researchers used two genome language models, Evo 1 and Evo 2, to write complete bacteriophage genomes from nothing but a design prompt. Of 285 designs they synthesised in the lab, 16 produced viable viruses that infect and kill E. coli, and a cocktail of them cleared bacterial strains that had already grown resistant to the natural phage the designs were modelled on. The work appeared Thursday in Science, eleven months after it was first posted as a preprint.
Why it mattersThis is the first time generative AI has designed a complete working genome rather than a single protein, and it points at phage therapies that could be redesigned faster than bacteria evolve resistance. It also sharpened the biosecurity argument within hours: researchers at Johns Hopkins responded by calling for legally required screening of synthetic DNA orders instead of the voluntary checks the industry relies on today. The models were trained without human viral sequences, and every experiment used non-pathogenic lab strains.
▶ Related video: 121. AI-designed viruses
Read in the app — free, in 9 languages
Related stories
Machine learning read the shape of sick brain cells and picked out nine already-approved drugs that calmed them down2026-08-21Given four hours and a GPU to improve the way AI is trained, the best agent scored 0.25 out of 1 — and most never tried
2026-08-21The agent invented a second person to vouch for its code. A 24-year-old in Texas refused to believe either of them.
2026-08-21Pew put half a million web pages through a detector: a third of everything published since ChatGPT carries its marks
2026-08-21The FDA has cleared 1,357 AI medical devices. Three of them have been tested on whether patients live longer or better.
2026-08-20