A team at Stanford University has used generative artificial intelligence to create viral genomes that do not exist in nature. The work, reported in Science on Thursday and covered by the Financial Times, produced laboratory-designed bacteriophages and has sparked both biomedical interest and biosecurity concerns.
What was done?
The researchers developed a generative AI model called Evo 2, capable of designing complete genomes — the DNA-encoded genetic instructions of organisms. Using this software, the team created 16 synthetic bacteriophages. In lab tests, these engineered phages destroyed Escherichia coli more effectively than the natural starting phage, ΦX174.
For safety, the experiments used a non-pathogenic E. coli strain, and the model was not trained on data from viruses that infect animal or plant cells.
Why this matters for healthcare
Phage therapy uses specific viruses to target and eliminate pathogenic bacteria. Although practiced in parts of Eastern Europe for decades, phage therapy has not been widely adopted in Western medicine. With the rise of antibiotic-resistant pathogens, however, phage-based approaches are gaining renewed interest, and Evo 2–designed phages could become a valuable addition to the therapeutic toolkit, noted Samuel King, a member of the Stanford research team.
Access and researcher rationale
Evo 2 is open source and freely available. Brian Hie, the project lead, argued that as an academic laboratory they have a duty to make the technology broadly accessible to realize potential health benefits.
Adrian Woolfson, CEO of the DNA-synthesis firm Genyro, described the result as a “Wright brothers moment” for biology, suggesting that science is moving from relying solely on evolutionary products of the past toward intentionally designing future biological possibilities.
Biosecurity concerns and regulatory gaps
Alongside enthusiasm, biosecurity experts have raised warnings. Thomas Inglesby and Moritz Hanke of the Johns Hopkins Center for Health Security wrote in Science that existing regulatory frameworks are likely insufficient to oversee generative genomics. They argue the issue is no longer whether generative virus-genome design will be achieved, but whether society can build oversight systems that allow beneficial uses while effectively preventing serious misuse.
Limits and future outlook
The authors stress that while current work incorporated safety precautions, the technology could be extended to more complex organisms in the future and might eventually operate without requiring a natural genome as a starting point. Such developments would raise further ethical, legal, and security questions.
Conclusion
Stanford’s team achieved a notable advance in synthetic biology by using a generative AI to design 16 bacteriophages that showed improved activity against E. coli in the lab. The result could accelerate phage therapy development but also underscores urgent biosecurity and regulatory challenges that need addressing to prevent misuse.
Tags: artificial intelligence, antibiotic, virus, medicine, genetics, biotechnology, bacteria, DNA, generative AI, gene research
Note: An AI assistant contributed to the preparation of this article; the final content was edited and verified by our journalist.
