In brief
U.S. researchers have employed generative AI models to design 16 fully functional bacteriophage viruses that replicate and attack only bacteria, notably E. coli. The first successful whole-genome AI creation marks a significant step in synthetic biology, offering potential treatments for antibiotic-resistant infections. Despite this promise, experts warn that the same technology could be weaponised, prompting calls for tighter biosafety safeguards. The study showcases AI’s ability to predict biological sequences, yet underscores the delicate balance between innovation and security.
AI has created 16 brand‑new viruses that function in the lab and can replicate. U.S. scientists announced the first time a generative AI model designed an entire viral genome, producing bacteriophages that selectively infect bacteria and pose no risk to humans.
The breakthrough, labeled a “very significant turning point” in science, could herald a new era of treatments for infections that have become resistant to conventional antibiotics. Yet, experts are immediately warning that AI‑driven virus design raises urgent biosafety and biosecurity concerns.
These AI models—named Evo1 and Evo2—are trained on genetic codes from a wide spectrum of organisms, including viruses, bacteria, plants and humans. By learning the language of life, the models predict nucleotide sequences that assemble into viable viral genomes.
From the 302 top AI designs, 16 were synthesized in the lab. When placed on petri dishes containing E. coli, the new phages produced clear zones where the bacteria were destroyed. Professor Brian Hie of Stanford University described the moment as “new territory for us.”
While the viruses designed so far are bacteriophages—viruses that infect bacteria—they open a pathway toward therapies that could target resistant infections in humans and reduce reliance on antibiotics.
Experts such as Dr Thomas Inglesby and Dr Moritz Hanke from Johns Hopkins highlighted that the study underscores the crucial need for safety protocols. They argue that the technology “remains usable only if it is restrained to prevent serious harm.” Projects like this do, however, implement strict safeguards: training data exclude viruses that can infect complex organisms, and experiments are confined to secure laboratories.
The implications stretch beyond phage therapy. Professor Marc Güell of Pompeu Fabra University praised the study as “the first time in history we can design biology on a computer,” envisioning future uses such as engineered enzymes for treating genetic disorders or novel antibodies for immunotherapy.
On the flip side, studies of synthetic biology raise ethical questions about the creation of artificial life. Although viruses are not alive, their genomes are a stepping-stone to AI-designed organisms that might overcome many biological limits. This dual nature of AI-driven genome design—potential to solve urgent health problems and to produce dangerous agents—makes the scientific community and regulatory bodies work hand‑in‑hand to balance progress with precaution.
