Researchers Have Generated Functional Phages Using AI

Scientists designed synthetic bacteriophages capable of infecting drug-resistant bacteria by training a model on DNA.

Updated on Sept. 18, 2026 in Life Sciences

A close-up view of a laboratory glass vial filled with blue-tinted liquid and tiny helical micro-structures, representing synthetic biological engineering. — Researchers from Stanford University and the Broad Institute have successfully designed 16 functional, synthetic bacteriophages using an AI model trained on genomic DNA sequences. AI Illustration. Upload story photo >

Live Poll

Do you believe using artificial intelligence to create new biological organisms is a positive development?

Researchers from Stanford University and the Broad Institute have created 16 functional bacteriophages using an AI model trained on genomic sequences. This research-stage development produced synthetic viruses capable of infecting E. coli strains that are resistant to natural phage counterparts.

Why it matters

By enabling the rapid design of novel viral cocktails, this AI-driven approach offers a potential path to overcoming the increasing threat of bacterial resistance to traditional phage therapies. The method allows for the creation of genetic variations that naturally occurring phages may not possess.

Researchers synthesized 285 AI-generated genomes, with 16 successfully demonstrating independent infection and survival. These synthetic sequences maintained a 93% to 98% identity to existing natural phages, while some variants differed by more than 5% from their nearest natural relatives.

The players

Stanford University

A leading academic research institution focused on advanced genomic engineering and artificial intelligence integration.

Broad Institute

A collaborative research center between MIT and Harvard specializing in large-scale genomic discovery and synthetic biology.

The details

The team utilized Evo, an AI model trained on the DNA of small bacteriophages—viruses that exclusively target bacteria. The model was prompted with the sequence of the phage ΦX174 as a template, generating variations that were computationally screened for size, gene organization, and host specificity. Successful candidates were then chemically synthesized to verify if they could infect E. coli, effectively bypassing the resistance mechanisms evolved against the template virus.

1915: Frederick Twort observed the first evidence of bacteria-destroying clear patches.
1917: Félix d'Hérelle formally identified the bacteriophage as an infectious agent.
Early 1970s: DNA sequencing technology enabled the precise reading of phage genome sequences.
September 18, 2026: Researchers published findings on the creation of functional AI-generated phages.

The Tech Race

This development signals a shift in the field of bacteriophage therapy from natural prospecting to generative design. It accelerates the timeline for creating bespoke antimicrobial agents, moving the race toward custom-designed viral cocktails that can specifically target resistant pathogens.

While this remains a research-stage development, the success demonstrates a workflow that could eventually lead to more effective treatments for antibiotic-resistant bacterial infections. Future clinical timelines depend on subsequent validation of these synthetic phage cocktails in broader bacterial populations.

The takeaway

The successful synthesis of functional phages via AI confirms that generative models can effectively navigate complex genomic space to design new biological tools. Interested researchers should monitor upcoming peer-reviewed publications for experimental data on the efficacy of these cocktails against broader, more diverse bacterial strains.