Researchers at Stanford University and the non-profit Arc Institute in Palo Alto, California, have successfully used artificial intelligence to design and generate complete genomes for functional viruses. These AI-designed phages, created by two AI models called Evo 1 and Evo 2, were trained on approximately 2 million bacteriophage genomes and were able to propose new genetic codes. Of the 302 AI-generated genome designs chemically printed as DNA strands and tested, 16 proved effective at replicating and killing E. coli bacteria.
This breakthrough marks the first time generative AI has been used to design a complete, functional genome. While the viruses created specifically target bacteria and pose no threat to humans, the development has been labeled a "very significant turning point" in science. Experts acknowledge the potential to revolutionize medicine by developing new drugs, therapies, and potentially more effective gene therapy vectors. For instance, the technology could accelerate research into artificially engineered cells and offer new approaches to treating antibiotic-resistant bacterial infections through phage therapy, with some AI-designed phages killing E. coli faster than naturally occurring ones.
However, the advancement also raises urgent biosafety and biosecurity questions. Dr. Thomas Inglesby and Dr. Moritz Hanke from Johns Hopkins University's Center for Health Security emphasized that it's no longer a matter of "whether generative viral genome design will exist" but how to use it responsibly without "enabling serious harm." The Stanford researchers explicitly avoided training their AI on human pathogens to mitigate risks, and the work was conducted in secure laboratories with phages that do not infect people. Despite these precautions, concerns remain about the potential for other scientists to misuse the methods to create more lethal human pathogens. The genetic code for the AI-designed phages is around 5,400 base pairs, significantly smaller than the 500,000 base pairs of the smallest living cell, indicating that creating larger, more complex organisms with AI is still a distant challenge.
