Scientists use AI to create synthetic bacterial viruses

Researchers at the Arc Institute and Stanford have used AI models Evo 1 and Evo 2 to design entirely new viruses capable of infecting and reproducing inside bacteria. These genome language models function similarly to LLMs but are trained on trillions of nucleotides, learning the structure and rules of genetic code. For safety, the team restricted training to viruses related only to Phi X-174, which infects E. coli and poses no threat to humans or other organisms. Evo generated around 700,000 potential viral genomes, from which scientists selected 285 candidates to synthesize and test. Sixteen of these AI-designed viruses successfully replicated, with some outperforming the original Phi X-174. While the research could revolutionize gene therapy and biotechnology, it also highlights serious risks, as future misuse could enable the creation of harmful or weaponized pathogens. The study underscores the urgent need for stronger regulations and safeguards as AI moves from describing biological threats to designing them.

Read the full story on Engadget →