Key Takeaway:
- AI created 16 new functional viruses that do not exist in nature, marking the first successful AI-designed viral genomes.
- The engineered viruses infect bacteria, not humans, and could help develop new treatments against antibiotic-resistant infections.
- Experts warn AI genome design outpaces regulation, raising urgent biosafety and biosecurity concerns.
Scientists at Stanford University and the Arc Institute have used artificial intelligence to design 16 new bacteria-infecting viruses from scratch, marking the first successful AI Virus genomes generated while prompting both medical optimism and biosecurity concerns.
AI Designs functional viruses for the first time
A study published Thursday in the journal Science reports that researchers successfully created 16 functional viruses that do not exist in nature using an artificial intelligence model trained on millions of genetic sequences.
The research team developed an AI system called Evo, which learned patterns from genetic data collected across all domains of life. The model generated thousands of possible viral genomes designed to infect E. coli bacteria.
Scientists synthesized and tested about 300 of those AI-designed genomes in laboratory conditions. Sixteen proved capable of functioning as viruses, making the study the first demonstration of AI creating complete, working viral genomes from scratch.
Researchers said the work represents a major advance in synthetic biology and AI-assisted genetic design, a fast-moving field covered in this overview of AI’s expanding role in biology.
Researchers target antibiotic resistance with engineered phages
The viruses created in the study were designed exclusively to infect bacteria, not humans. Researchers intentionally excluded datasets containing human pathogens while training the AI model.
According to the study’s authors, the technology could support the development of bacteriophage therapies that adapt to drug-resistant bacteria, offering a possible new approach to combating antibiotic-resistant infections and easing concerns tied to any lab-made AI Virus strain.
“The viruses were engineered specifically to target bacteria, not people,” the researchers wrote in the study, noting that the AI-generated viruses are not capable of infecting humans.
Scientists said AI-generated phages could eventually become more resilient than naturally occurring viruses, improving treatments against bacteria that evolve resistance to existing therapies.
Experts warn governance lags behind AI capability
The findings also triggered warnings from biosecurity experts about the rapid advancement of AI-powered genome design.
In a companion article published in Science, Thomas Inglesby and Moritz Hanke of the Johns Hopkins Center for Health Security wrote that the research 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,” Inglesby and Hanke wrote.
The experts argued that discussions should now focus on preventing misuse rather than questioning whether the technology will emerge. They cautioned against pursuing AI-designed viruses capable of causing disease and called for stronger governance frameworks before such capabilities become more widespread, according to CNN’s coverage of the study.
For now, the researchers emphasized that the work remains limited to bacteria-infecting viruses developed under tightly controlled laboratory conditions and does not involve viruses that infect people, keeping this early-stage AI Virus research confined to the lab for now.
The study appears as scientists increasingly explore AI applications in biology, including drug discovery, protein engineering and genetic research, while policymakers debate safeguards for rapidly advancing technologies.
