Researchers at Stanford University have used artificial intelligence to design brand new viruses. This is the first time a complete virus genome has been built entirely by AI.
The viruses are called bacteriophages. These viruses infect bacteria, not humans or animals.
The AI systems used for the project are called Evo1 and Evo2. They work in a similar way to language models like ChatGPT, but instead of predicting words, they predict genetic code.
The models were trained on genetic sequences taken from viruses, bacteria, plants, and people. Scientists then refined them to focus on designing bacteriophages that target specific types of bacteria.
How The Experiment Worked
The research team generated 302 possible virus designs using the AI models. They built and tested each design in a lab setting.
Out of the 302 designs, 16 worked. These viruses were able to kill E. coli bacteria during petri dish tests.
PhD student Samuel King said the team noticed the viruses working during early morning testing. Clear spots appeared on the dishes where the bacteria had been destroyed.
Assistant professor Brian Hie said the room broke into applause once the results were shared with the wider team. He described the project as new territory for AI research.
Why This Could Help Medicine
New bacteriophages could offer another way to treat infections that no longer respond to antibiotics. Antibiotic resistance is a growing problem for doctors around the world.
Hie said the technology has the potential to improve human health by helping create new drugs and therapies. He described the work as an early step rather than a finished cure.
The genetic code for the new phages is about 5,400 base pairs long. The human genome, by comparison, contains around three billion base pairs.
Hie said building simple living organisms with AI would be hard but not impossible. He said the team wants to work toward that goal in the future.
Other scientists praised the study. Professor Marc Güell from Pompeu Fabra University in Spain called it a turning point for designing biology on a computer.
Professor Patrick Cai from the Manchester Institute of Biotechnology called the work an important milestone. He said it shows AI models are starting to learn design patterns found in nature.
Not everyone focused only on the positives. Biosecurity experts Dr Thomas Inglesby and Dr Moritz Hanke from Johns Hopkins University said the findings raise safety and security questions.
Writing in the journal Science, they said the real question is no longer whether AI can design viruses. Instead, they said the question is whether the tool can be used without allowing harm.
The Stanford team said it took safety steps during the research. They excluded viruses that could infect complex organisms from the training data and carried out testing in a secure lab.
The researchers worked only with bacteriophages, which do not infect people. They said this choice helped limit risk during the project.
