Scientists have synthesized bacteriophages whose genomes were generated by artificial intelligence, according to a study published in Science on August 6, 2026. The viruses, which infect bacteria and are used to treat persistent infections, were tested in laboratory dishes and killed Escherichia coli strains resistant to naturally occurring bacteriophages.

Dr Brian Hie, a chemical engineer at Stanford University, employed genome language models to design functional bacteriophage genomes. The AI models, trained on genetic data from two million bacteriophages, excluded sequences that could produce plant, human, or animal viruses to reduce safety risks. From thousands of proposed designs, the researchers selected nearly 300 for laboratory synthesis; only 16 proved viable. A mixture of these phages rapidly overcame resistance in two E coli strains.

The researchers noted that the approach could expand phage therapy and broaden biotechnological tools, but they also highlighted biosafety, biocontainment and biosecurity concerns. They called for consultation with safety and security professionals throughout genome design.

Prof Tom Inglesby and Dr Moritz Hanke of the Center for Health Security at Johns Hopkins University wrote that, while the work holds promise for life‑science applications, it raises urgent questions about governance. They warned that the ability to compose viral genomes using generative AI now exists without established regulatory frameworks.

Tom Ellis, a professor of synthetic genome engineering at Imperial College London, described the achievement as impressive but noted that the genome created is among the simplest to construct. He added that designing more complex or pathogenic genomes would be more challenging, and that current threats from AI‑generated viruses may be overstated compared with modifying existing pathogens.

Dr Filippa Lentzos of King’s College London emphasized the need for layered governance, including safeguards at the model‑development stage, responsible research review, synthesis screening, and established laboratory biosafety measures.

Bacteriophage genomes are small, but the study demonstrated that generative AI can produce functional viral genomes. The authors indicated that applying the same methods to pathogens capable of infecting humans, animals or plants remains uncertain, and they advised against pursuing such work without robust oversight.