Scientists have used AI to design new kinds of viruses for the first time, marking a major advance in biotechnology while prompting fresh concerns about how the same tools could be misused.
The newly created viruses are bacteriophages, also known as phages, which infect bacteria rather than people, animals or plants. Researchers from Stanford University and the Arc Institute designed them to target strains of E. coli in laboratory experiments. The 16 functioning viruses produced during the study are not considered a danger to humans.
The team used genomic AI models called Evo1 and Evo2. Similar to how language models learn patterns in written text, these systems were trained to recognize patterns within genetic code and generate new DNA sequences.
After the models proposed thousands of possible genomes, researchers selected almost 300 for laboratory production. Most failed to function, but 16 successfully reproduced and infected bacteria. This showed that AI could generate the complete genetic instructions needed for a working virus rather than merely suggesting changes to an existing organism.
The experiment may have important medical applications. Bacteriophages are already being investigated as treatments for infections that no longer respond to antibiotics. In the study, combinations of the AI-designed viruses destroyed two E. coli strains that had developed resistance to naturally occurring phages.
Researchers believe the technology could eventually help scientists create treatments tailored to particular bacterial infections. New phages might also be generated more quickly as bacteria evolve resistance, potentially strengthening efforts to tackle the growing problem of antimicrobial resistance.
However, the ability to write functioning viral genomes has also intensified the debate over biological security. The researchers excluded genetic information from viruses that infect humans, animals and plants from the models’ training data. They also conducted their experiments using bacteria that were not harmful to people.
Experts nevertheless warned that future AI systems could become more powerful, while other developers might not adopt the same precautions. They called for safeguards at several stages, including restrictions on sensitive training data, independent reviews of research projects, laboratory controls and screening by companies that manufacture synthetic DNA.
Some scientists have cautioned against exaggerating the immediate threat. The bacteriophages used in the experiment have extremely small genomes, and designing a complex virus capable of infecting humans would be substantially harder. Modifying an existing pathogen may currently present a more realistic risk than creating one entirely through AI.
The research therefore represents both a promising medical breakthrough and an early warning. AI could help develop new weapons against dangerous bacteria, but its growing ability to design biological systems will require oversight that advances alongside the technology.
Add Comment