Scientists have created the first viruses designed by artificial intelligence, marking a breakthrough that could lead to new medical treatments while raising important safety concerns. Bacteriophages infect only bacteria and already help treat stubborn bacterial infections.
Using genome language models, the genetic counterpart of the AI systems behind chatbots, Dr. Brian Hie of Stanford University designed complete bacteriophage genomes. Scientists built the AI-designed viruses in the laboratory, and they successfully destroyed E. coli bacteria that had resisted naturally occurring bacteriophages.
AI-Designed Viral Genomes Raise Promise and Biosafety Concerns
Researchers say the ability to rapidly design viral genomes and tailor them to target specific bacteria could revolutionize phage therapy and expand biotechnology. However, they warn that the breakthrough also raises major biosafety and biosecurity concerns, urging scientists to work closely with safety and security experts when creating whole genomes.
Using AI models called Evo1 and Evo2, trained on genetic data from two million bacteriophages, the team generated thousands of candidate viral genomes. To reduce risk, researchers trained the models only on bacteriophages and excluded viruses that infect humans, animals, or plants. Of nearly 300 genomes built in the lab, only 16 produced viable viruses, but together they successfully defeated resistant E. coli strains.
Experts Urge Strict Limits on AI-Designed Pathogens
Experts at Johns Hopkins University said the study demonstrates that generative AI can now create functional viral genomes, even if only for simple viruses. They argued that scientists should avoid researching pathogens that can infect humans, animals, or plants because such work could produce organisms that current medical defenses cannot contain.
Other scientists noted that bacteriophage genomes are among the simplest to engineer, making more complex viruses far harder to design. They emphasized that the greater immediate risk comes from modifying existing pathogens rather than creating entirely new ones with AI. To reduce future dangers, they recommend layered safeguards, including restricted access to sensitive genetic data, DNA synthesis screening, responsible research oversight, and strong laboratory biosafety measures.
Read the original article on: the guardian
Read more:A key meat protein has been successfully produced in lettuce for the first time
