Researchers at Stanford University have successfully used artificial intelligence to design functional, replicating viruses. This marks a first in scientific history where generative AI models created complete genomes that operate within a laboratory setting. The team focused their efforts on bacteriophages, which are viruses that specifically target and kill bacteria, rather than those that affect humans.
The AI models, identified as Evo1 and Evo2, function similarly to language models. Instead of processing text, they analyze genetic code. The researchers trained these models on a wide array of viral, plant, and human genetic sequences. After generating hundreds of potential designs, the team synthesized 302 of them. Of those, 16 showed effectiveness at eliminating E. coli bacteria in a controlled environment.
This development serves as a major milestone for synthetic biology. Scientists believe this capability could eventually lead to new treatments for infections that current antibiotics cannot clear. By programming biological structures on a computer, researchers hope to create enzymes for genetic disorders or new antibodies for immunotherapy. The ability to write genetic code opens a new field of study in genomic design.
However, the breakthrough also prompts serious safety and security questions. Experts from the Johns Hopkins Center for Health Security noted that the existence of generative viral design technology presents an urgent need for oversight. While the Stanford team limited their project to phages that pose no threat to humans, critics argue that the same methods could potentially be misused to create harmful diseases.
Looking ahead, the researchers acknowledge that moving from viral genomes to more complex living organisms remains a massive hurdle. Phage genomes contain approximately 5,400 base pairs, whereas even the simplest living cell requires hundreds of thousands. While the team remains interested in exploring these limits, the immediate priority remains the refinement of phage therapy for medical applications.