Researchers have used artificial intelligence to identify and create functioning viruses not seen in nature for the first time.
The research, announced in the journal Science on Thursday, “expands what synthetic genomics can achieve,” according to its authors from Stanford University and the California-based Arc Institute.
Stanford chemical engineer Dr. Brian Hie said his team’s work could help inform future medicines aimed at treating drug-resistant bacteria using bacteriophages, viruses that infect and kill bacteria.
“If the bacteria gain resistance to a single phage, it’s game over for the medication,” Hie told the Stanford website on Thursday. “But if you have multiple genetically distinct phages in a mixture, it would be harder for the bacteria to develop resistance to the entire cocktail.”
Others were alarmed that governments are not equipped to handle these rapidly advancing technologies, which could fall into the hands of bad actors.
“Although this is promising for life sciences applications, it also raises urgent biosafety and biosecurity questions,” experts from the Center for Health Security at Johns Hopkins University wrote in a commentary accompanying the research in Science. “The ability to compose viral genomes using generative AI now exists; the governance to safely steer it does not.”
To build their novel virus, the researchers trained an open-source AI model known as Evo on the genetic sequences of millions of animals, plants, microbes and viruses, excluding data from viruses that infect humans.
They then homed in on the virus ΦX174, a well-studied bacteriophage that infects only E. coli, training the AI on its genome and those of similar viruses.
Using this information, the model generated 700,000 potential new viruses, 285 of which the researchers converted into DNA molecules. Sixteen produced viable viruses when inserted into bacteria.
The newly created viruses showed promise in overcoming resistance from their hosts, suggesting similar methods could be used to develop medication targeting tuberculosis and methicillin-resistant Staphylococcus aureus, according to the researchers.
“We have a proof of concept in the paper, where we show that this cocktail of 16 phages rapidly overcomes resistance in E. coli that is immune to native ΦX174,” Hie told the Stanford website.
AI systems, with their ability to detect patterns in large datasets, have proven adept at studying lengthy sequences of DNA and proteins.
In 2024, researchers at Google DeepMind won a Nobel Prize for AlphaFold, which predicts the complex three-dimensional structure of proteins.
Tom Ellis, a professor of synthetic genome engineering at Imperial College London, told The Guardian that despite fears about how virus-creating AI could be used, existing pathogens and technologies to tweak viruses are more concerning.
“The threat from full AI design and writing of a genome of a virus or bacteria is very overblown when we consider that just taking existing pathogens and making gain-of-function changes to their genomes is so much easier and much more likely to be a real pathogenic threat,” he said.
Concern over gain-of-function research, science which alters viruses to change their properties and study their evolution, prompted the National Institutes of Health to announce a ban on the practice last month.
“This groundbreaking government-wide policy establishes, for the first time, an explicit prohibition on dangerous gain-of-function research which can significantly endanger American lives and pose substantial risk to our Nation’s security,” NIH Director Dr. Jay Bhattacharya wrote in a statement.
Many figures in the Trump administration and Republican party believe a leak from a virology lab in China caused the Covid pandemic, though the World Health Organization stated in a 2025 report that the disease likely spilled over from animals to humans.
Concerns abound that powerful and easily accessible AI models could create runaway threats, ranging from pathogens to cyberattacks.
Last month, ChatGPT maker OpenAI announced that its experimental systems broke out of its safeguards and hacked a fellow AI company.