Scientists from Stanford University in the United States have used generative artificial intelligence for the first time to create viruses capable of killing treatment-resistant bacteria.
A research team led by chemical engineer Brian Hie applied a genome language model to design new bacteriophages, which are viruses that infect only bacteria. The study, published in the journal Science, noted that natural phages have lost effectiveness as dangerous infections such as E. coli developed resistance.
The AI-created viruses showed the ability not just to withstand pathogens, but to destroy them. The scientists said the discovery could change not only medical research, but the entire healthcare system.
Biosecurity Risks and Early Stage Research
The authors of the study also warned of potential risks. To prevent the technology from being used for harm, the researchers did not include pathogens dangerous to humans or animals in the project, and they urged all followers to consult with safety specialists at all stages.
The researchers specified that their work remains at an early stage. They added that no current AI model has enough power to analyze human or animal biochemistry.
Expert Warnings on Artificial Intelligence Threats
In an accompanying article in Science, Professor Tom Inglesby and Dr. Moritz Hanke from the Johns Hopkins Center for Health expressed even greater concern about the breakthrough. They wrote that while the work is promising for life sciences, it poses urgent biosecurity questions because the ability to construct viral genomes using generative AI now exists, while governance to safely control the process does not.
Earlier, Artem Geller, an expert with the Commission for the Development of the Information Society in the Federation Council and developer of the Kremlin website, stated that the rapid spread of AI-based technologies carries serious threats, including increased data leaks and cyberattacks. He said neural networks should be treated like new employees whose access to sensitive information must be strictly limited.
