Researchers used AI to design new viruses that can replicate and kill bacteria in the lab
Artificial intelligence has been used to design brand new viruses that can replicate in a laboratory, marking what scientists say is a major turning point in synthetic biology.
Researchers in the United States created 16 novel viruses designed to infect bacteria. The viruses do not pose a threat to humans, but the breakthrough has raised both excitement and concern because it shows that AI can now help design whole genomes.
The AI virus design work was carried out by researchers at Stanford University. They used advanced models to generate new genetic sequences for bacteriophages, a type of virus that infects bacteria rather than people.
The team said this is the first time whole genomes have been successfully designed by generative AI. That matters because designing a complete, functioning virus is far more complex than using AI to suggest a new drug or antibiotic.
Brian Hie, an assistant professor at Stanford, described the research as new territory. He said generative AI had now moved into designing something that can replicate and function inside cells.
The technology works in a similar broad way to language models, which predict sequences of words. In this case, however, the AI models were learning patterns in genetic code rather than human language.
The models, known as Evo1 and Evo2, were trained on genetic information from viruses, bacteria, plants and people. They were then refined to focus on designing bacteriophages.
From the AI-generated options, the researchers selected 302 promising designs and made them in the lab. Sixteen of those proved capable of killing E. coli bacteria.
The moment the team saw the results was dramatic. Clear spots appeared on dishes where bacteria had been growing, showing that the new phages were working. According to the researchers, the discovery prompted excitement across the lab.
The potential medical value is significant. Bacteriophages are already being studied as possible weapons against antibiotic-resistant infections. As more bacteria become resistant to existing drugs, phage therapy is increasingly seen as one possible route for treating infections that no longer respond to antibiotics.
Supporters of the work believe AI could speed up that process by helping scientists design new biological tools more quickly. In the future, similar methods could help create treatments, enzymes, antibodies or other therapies aimed at disease.
But the same breakthrough also raises serious safety questions.
Experts warned that if AI can design useful viruses, it could also be misused to create dangerous ones. In a commentary linked to the research, specialists from the Johns Hopkins Centre for Health Security said generative viral genome design now raises urgent biosafety and biosecurity concerns.
They warned that the issue is no longer whether this kind of technology will exist, but whether it can be used safely without enabling serious harm.
The Stanford team said it took steps to reduce risk. They did not train the system on viruses that infect complex organisms, and the work focused on phages rather than viruses that infect humans. The experiments were also carried out in a secure laboratory setting.
There are still major limits. Viruses are not alive, and creating a living organism would be far more difficult. The phage genome used in this work is about 5,400 genetic letters long, while even the smallest living cell has a genome hundreds of thousands of letters long. The human genome is around three billion letters.
Even so, scientists say the direction of travel is clear. AI is beginning to move from analysing biology to writing it.
That could bring huge medical benefits. It could also force governments, laboratories and technology companies to decide how powerful biology-design tools should be controlled.
For now, the breakthrough is both promising and unsettling. AI has designed viruses that work in the lab. The next question is how science keeps that power useful, safe and tightly governed.