---
title: AI Designed the Genomes of Sixteen Working Viruses From Scratch
description: Stanford and Arc Institute researchers designed 16 working bacteriophages with genome language models, published in Science alongside a call for oversight.
author: Darie Nani (Editor-in-Chief)
updated: 2026-08-06T21:25:22.466Z
canonical: https://www.sovereignmagazine.com/article/ai-designed-bacteriophage-genomes-science
image: https://cdn.nanimediahouse.com/samuel-king-phage-genome-design-113816.webp
categories: Science & Tech
content_type: News
region: Global
publication: Sovereign Magazine
schema_type: Article
---

Researchers at Stanford University and the [Arc Institute](https://arcinstitute.org/news/hie-king-first-synthetic-phage) used AI to design complete genomes for bacteriophages, the viruses that infect bacteria, then made the designs in the lab and tested them. Their models generated hundreds of candidate genomes based on PhiX174, a small and well studied phage used as the template. Sixteen of those candidates turned out to be working viruses, with genetic sequences and structures that differed substantially from each other. The study, led by Samuel King, was published on 6 August in Science.

Advances in DNA sequencing and synthesis have made it increasingly possible to read and write whole genomes, but designing a functional one from scratch is extraordinarily difficult, because genes, regulatory sequences and other elements interact in complex ways.

## The Designs Came From the Team's Own Genome Language Models

The sequences came from Evo, a family of genome language models the same group had built, including the earlier Evo 1. King and his colleagues, with Brian Hie of Stanford and the Arc Institute as corresponding author, paired the generated sequences with computational biology and laboratory screening rather than trusting the prediction alone. The authors describe the result as a step toward designing entire biological systems rather than single genes or small genetic systems.

## Combinations of the New Phages Overcame a Resistant E. coli

Several of the designed phages performed comparably to their naturally occurring relatives. In some cases combinations of them overcame resistance in two strains of E. coli that had resisted PhiX174-like phages. Phages are already used as treatments for bacterial infections, and the authors say AI-guided genome design could eventually produce more durable phage therapies. Eric Topol, the physician-scientist, wrote in a post on X on 6 August that generative AI in biology has moved from designing proteins to designing viral genomes, and that the approach has potential for making phage to counter antimicrobial resistance.

## The Authors Asked for Safety Review Alongside the Work

King and colleagues write that designing and synthesizing functional AI-generated genomes raises biosafety and biosecurity concerns, and needs expert oversight throughout the design process.

> "Groups conducting future whole-genome design work should consult both safety and security professionals throughout the project lifecycle."
> — Samuel King and colleagues, writing in Science

They argue existing safety frameworks can be adapted to generative genomics, and that protections built into the models themselves, such as excluding sensitive viral sequences from training data, could add another layer.

## A Companion Article Says the Open Question Is Oversight

In a Perspective published alongside the paper, Thomas Inglesby and Moritz Hanke of the Johns Hopkins University School of Public Health write that the authors “engage with biosafety and biosecurity questions more deliberately than most developers of powerful biological AI models”. They add: “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.”

## FAQ

**Q: What is a bacteriophage?**
A virus that infects bacteria rather than human or animal cells. Phages are used in research and, in phage therapy, as treatments for bacterial infections.

**Q: Can these AI-designed viruses infect people?**
No. The team designed bacteriophages, using PhiX174, a phage that infects E. coli, as the template. Nothing in the study describes viruses that can infect human cells.

**Q: What is phage therapy used for?**
Treating bacterial infections. The authors say AI-guided genome design could eventually help produce more durable phage-based treatments.

**Q: What safeguards do the researchers propose?**
Consulting safety and security professionals throughout a project, adapting existing biosafety frameworks to generative genomics, and protections inside the models themselves, such as leaving sensitive viral sequences out of training data.