AI-Designed Bacteriophages Show Promise Against Antibiotic-Resistant Bacteria in Lab Tests

First AI-created viruses target antibiotic-resistant bacteria in lab trials, offering potential for cheaper antimicrobial solutions while explicitly avoiding human pathogens.

Researchers at Stanford University and Johns Hopkins University created the world’s first artificial intelligence-designed viruses—specifically bacteriophages that infect only bacteria—in a study published in *Science* on August 6, 2026. Using genome language models (Evo1 and Evo2) trained on genetic data from 2 million bacteriophages, they generated nearly 300 candidate genomes. After lab testing, 16 proved viable, and a cocktail of these phages successfully overcame antibiotic resistance in two strains of *E. coli* that had previously evaded natural bacteriophages.

The work represents a controlled, scalable approach to engineering microbes with targeted effects. Crucially, the researchers intentionally excluded genetic sequences for human-pathogenic viruses, animal infections, or plant diseases during training. They explicitly state this method cannot be used to design viruses affecting humans, and they urge ongoing safety consultations throughout the process.

This breakthrough moves health abundance by potentially enabling cheaper, more precise antimicrobial treatments for antibiotic-resistant infections—a growing global crisis. If scaled, AI-designed bacteriophages could reduce reliance on costly antibiotics while avoiding harmful side effects. However, the current work remains confined to lab settings, and whether the same approach applies to other viruses beyond bacteriophages remains unverified. The next critical step is determining how these engineered microbes can be safely deployed without triggering unintended ecological or health consequences.

Source: Guardian