AI-designed bacteriophages show promise for fighting antibiotic resistance, raising biosecurity concerns

Scientists using artificial intelligence have designed functioning viruses for the first time, marking a significant breakthrough that offers both medical promise and urgent safety questions. Researchers led by Stanford's Dr Brian Hie employed genome language models — AI systems trained on genetic data from 2 million bacteriophages — to create new viral genomes targeting antibiotic-resistant bacteria. In laboratory tests, a combination of 16 AI-designed bacteriophages successfully defeated E coli strains that resisted natural bacteriophages, demonstrating the technology's potential in phage therapy. However, experts including Johns Hopkins Center for Health Security researchers Tom Inglesby and Moritz Hanke warned that while the breakthrough is promising, the governance frameworks to safely manage such capabilities do not yet exist. The researchers deliberately excluded genetic data for plant, human, and animal-infecting viruses from the AI training to limit risks, but they emphasized that researchers working with pathogenic genomes should not pursue similar design approaches without strict safety protocols and consultation with biosecurity professionals.