Generative AI has moved from reading genetic code to designing complete viral genomes that work in the lab.
Researchers at Stanford University and the Arc Institute used the Evo 1 and Evo 2 AI models to generate bacteriophage genomes designed to infect E. coli. Of roughly 300 designs synthesized and tested, 16 produced functional viruses, according to research published in Science. The viruses infect bacteria, not humans.
Unlike earlier efforts that copied or modified existing viruses, the AI generated entirely new viral genomes. The researchers trained the models on genetic sequences from millions of organisms and then asked them to produce genomes capable of infecting E. coli bacteria. Of roughly 300 AI-generated genomes synthesized and tested in the lab, 16 turned out to be fully functional viruses.
Why researchers are excited
The new viruses were able to kill E. coli strains, including some that had developed resistance to natural bacteriophages. Scientists say that ability could eventually help develop new treatments for antibiotic-resistant infections, one of the world’s growing public health threats.
Stanford computational biologist Brian Hie, one of the study’s authors, told the BBC that the experiment pushed generative AI into a new level of biological complexity. He described designing something capable of replication as “new territory for us.”
The safety debate begins
The researchers intentionally excluded viruses that infect humans, animals, plants and fungi from the AI’s training data and conducted the work in a secure laboratory. Even so, biosecurity experts say the study demonstrates that generative AI can be used to design genome sequences that produce functional viruses.
In a commentary published alongside the study, Dr. Thomas Inglesby and Dr. Moritz Hanke of the Johns Hopkins Center for Health Security warned: “Although this is promising for life sciences applications, it also raises urgent biosafety and biosecurity questions. The ability to compose viral genomes using generative AI now exists; the governance to safely steer it does not.”
A turning point for biotech
This research matters because it suggests AI is moving beyond analyzing biology and toward designing it. If the same approach eventually works on larger and more complex genomes, scientists could develop custom viruses for gene therapy, personalized treatments for bacterial infections, or other biotechnology tools that are difficult to engineer today.
The immediate commercial opportunity is likely to be in phage therapy and synthetic biology rather than human viruses, which remain far more complex than the bacteriophages used in the study. The process was also inefficient, with only 16 successful viruses emerging from hundreds of tested designs.
What changes after this study is not that dangerous human viruses can suddenly be created by AI, but that researchers have shown AI can design a complete viral genome that works in the real world. That shifts the conversation from whether AI-generated viral genomes are possible to how governments, laboratories, and DNA synthesis companies should manage the technology before it becomes far more powerful.
Other News: For other smartwatch news, see how Samsung Galaxy Watch models can display glucose readings through compatible CGM apps, watch faces, and connected Android phones.