Stanford researchers use generative AI to design functional new viruses
Researchers from Stanford University and the Arc Institute have successfully used generative AI models, known as Evo1 and Evo2, to design complete and functional viral genomes from scratch. This achievement, published in the journal *Science*, marks the first time AI has been used to create entirely new, self-replicating viral structures.
The study focused on designing bacteriophages—viruses that specifically infect bacteria—using the Phi X-174 virus as a template. Out of approximately 300 chemically synthesized AI-designed genomes tested in laboratory settings, 16 resulted in viable bacteriophages. These new viruses were found to be effective at infecting and destroying *E. coli* bacteria, even demonstrating an ability to overcome certain bacterial resistances more effectively than naturally occurring phages.
While the research offers significant medical potential, particularly for developing new therapies against antibiotic-resistant bacteria, it has also triggered urgent biosecurity warnings. Experts caution that the ability to compose viral genomes via generative AI could potentially be misused to create dangerous pathogens. Although the current research utilized non-pathogenic strains and models trained to avoid human or animal hosts, specialists emphasize that existing regulatory frameworks and governance may be insufficient to manage the rapid advancement of such generative biology tools.
Entities: Arc Institute · Brian Hie · E. coli · Escherichia coli · Evo 1 · Evo 2 · Evo1 · Evo2 · Science · Stanford University
Claims
What the coverage asserts, and how well corroborated each claim is across sources.
- [● 9 SOURCES] The 16 functional viruses were designed to infect and destroy E. coli bacteria. (Stanford University researchers)
- [● 2 SOURCES] Current governance and regulatory frameworks may be insufficient to manage rapidly developing generative biology tools. (Experts)
- [● 12 SOURCES] Out of nearly 300 chemically synthesized AI-designed genomes, 16 resulted in viable bacteriophages. (Stanford University researchers)
- [● 13 SOURCES] Researchers at Stanford University used generative AI to design complete, functional viral genomes from scratch. (Stanford University researchers)
- [● 10 SOURCES] The AI-designed phages were effective at overcoming bacterial resistance in E. coli. (Stanford University researchers)
- [● 11 SOURCES] Experts warn that the ability to compose viral genomes via generative AI raises urgent biosecurity and bioprotection concerns. (Experts)
- [● 7 SOURCES] The bacteriophages created do not pose a threat to humans or animals. (Stanford University researchers)
- [● 7 SOURCES] The AI models used for the research are known as Evo1 and Evo2. (Stanford University researchers)