Researchers at Stanford University and the Arc Institute have used artificial intelligence to generate entirely new virus structures that do not exist in nature, according to a study published in the journal Science. Scientists deployed an AI model named Evo 2 to analyze nine billion nucleotides across the PhiX174 genome and roughly 15,000 close relatives. The project aims to develop adaptive bacteriophages capable of targeting and destroying drug-resistant bacteria without harming human cells.
AI Generates Novel Bacteriophages to Fight Superbugs
Synthetic biology has long replicated known genomes, but the Evo 2 model takes a different approach by learning DNA structural patterns to write genetic recipes for entirely new viruses. According to reporting by The New York Times, researchers tested the AI-generated sequences by inserting them into E. coli bacteria that were resistant to PhiX174. Out of 700,000 computer-generated virus designs, scientists selected 285 sequences for chemical synthesis in the laboratory. Laboratory tests confirmed that 16 of these synthesized molecules successfully infected target bacteria while matching the resilience of natural viral strains.
Safety Guardrails and Laboratory Precautions
To prevent potential biological risks, the research team restricted the AI model to designing viruses that strictly target bacterial pathogens. According to Stanford University computational biologist and study co-author Brian Hie, the team wanted to remain extremely cautious during the investigation. The newly engineered bacteriophages replicate the behavior of natural PhiX174 viruses, which infect only microbial hosts and pose no threat to humans, animals, or fungi.
Did you know?
The Evo 2 model processed approximately nove biliões nucleotides from the PhiX174 genome and about 15 mil related sequences to learn natural DNA patterns before generating its novel virus designs.
Biosafety Debates and Institutional Responses
The breakthrough arrives amid heightened global scrutiny regarding laboratory-generated pathogens and biosecurity protocols. Moritz Hanke, a researcher at the Johns Hopkins Center for Health Security who did not participate in the study, told The New York Times that the researchers showed commendable restraint by limiting the scope of their AI model. However, Hanke also warned that governments and scientific organizations have been slow to establish comprehensive safeguards against the generation of dangerous biological agents.
As computational tools become more advanced in genomic design, scientific institutions continue to evaluate where to draw the line between therapeutic innovation and biosecurity oversight.
Frequently Asked Questions
Can these AI-generated viruses infect humans?
No. According to the study published in Science, the AI model was specifically trained to produce viral recipes that target only bacteria, mirroring natural bacteriophages like PhiX174 that do not affect humans, animals, or fungi.
What is the main goal of using AI to create new viruses?
Researchers aim to develop adaptive, resilient bacteriophage therapies to combat fast-evolving drug-resistant bacteria, commonly known as superbugs.
How many new virus designs did the AI model create?
The Evo 2 model generated 700,000 novel virus sequences, from which researchers selected 285 for chemical synthesis and laboratory testing.
Explore More: Want to stay updated on the latest breakthroughs in biotechnology and artificial intelligence? Subscribe to our daily newsletter for breaking science news, or share your thoughts on the ethics of synthetic biology in the comments below.
Worth a look