Researchers have used artificial intelligence to design complete, functional bacteriophage genomes and tested them against bacteria that had developed resistance to a natural bacteriophage, according to an Aug. 7 report. The study marks progress toward generative systems capable of engineering entire biological systems, while also raising biosafety and biosecurity concerns.
Thomas Inglesby and Moritz Hanke wrote in a related Perspective that the authors "engage with biosafety and biosecurity questions more deliberately than most developers of powerful biological AI models." Advances in DNA sequencing and synthesis have made it possible to read and write entire genomes, but designing a fully functional genome remains challenging due to the complex interactions between genes, regulatory sequences, and other elements.
Samuel King and colleagues introduced an approach combining Evo genomic language models—including Evo 1—computational biology, and experimental screening. Using this method with the well-studied ΦX174 bacteriophage as a model system, they computationally designed hundreds of candidate genomes. They experimentally identified 16 functional phages with genetic sequences differing substantially from each other. Some engineered phages performed similarly to naturally occurring relatives; notably, certain combinations overcame resistance in two strains of E. coli that resisted ΦX174-like phages.
The findings suggest that AI-guided generative genomics could eventually enable more durable phage-based therapies for bacterial infections. However, King et al. emphasized the need for expert oversight throughout the process: "Groups conducting future whole-genome design work should consult both safety and security professionals throughout the project lifecycle," they say.
They argue existing safety frameworks can be adapted for generative genomics applications while recommending additional safeguards, such as excluding sensitive viral sequences from training data at the model level. Inglesby and Hanke said, "The question is no longer whether generative viral genome design will exist. It is whether society can build oversight that allows its benefits to unfold while preventing it from enabling serious harm."
