Scientists report creating the first AI-designed bacteriophages by generating complete viral genomes in silico and then synthesizing and testing them in the laboratory. The work, published in *Science* and led by researchers at Stanford University, uses generative “genome language models” to write DNA sequences for viruses. In laboratory experiments, 16 AI-designed phages replicate and kill *E. coli*, including strains resistant to naturally occurring bacteriophages.
Across outlets, the broader implication is the potential to accelerate phage therapy by enabling rapid, customized design against specific bacterial infections, as antibiotic resistance becomes more widespread. Several reports describe the approach as a shift from designing individual genetic parts to designing whole, functional viral genomes that work inside cells. This is presented as a proof of concept for generative design in biology.
Other coverage emphasizes the risks and governance gaps. Multiple experts quoted in the reporting warn that the same capability to compose viral genomes could, in principle, be misused, and that current oversight may not keep pace with the technology. Suggested mitigation themes include layered safeguards, responsible review, and attention to controls around DNA synthesis and access to genetic data, alongside established lab biosafety and biosecurity practices.
