Researchers at Stanford University and the Arc Institute in Palo Alto have successfully used generative AI to design entirely novel, functional viruses. Detailed in a study published in Science, scientists used genome language models, Evo 1 and Evo 2, to generate genetic sequences capable of infecting and replicating inside bacteria.
Read: SpaceX shares plunge after first earnings release since IPO
Unlike large language models trained on human text, Evo was trained on trillions of nucleotides—the fundamental building blocks of DNA, to master the structural “grammar” of genetic code.
To conduct the study safely, researchers specialized Evo’s training on 15,000 viral genomes belonging to the bacteriophage family that infects E. coli. The team explicitly restricted the model’s design scope to exclude sequences capable of infecting humans, animals, plants, or fungi.
- Generation: Evo designed roughly 700,000 potential synthetic virus genomes.
- Synthesis: Scientists selected the 285 most promising candidates, chemically manufactured their DNA, and introduced them into host bacteria.
- Viability: 16 of the AI-designed candidates successfully produced fully functional viruses, demonstrating replication speeds equal to or faster than naturally occurring variants.
The breakthrough highlights the dual-use dilemma inherent in advanced biological AI. In controlled settings, AI-designed viruses could revolutionize biotechnology, accelerating the development of targeted gene therapies, precise tools for targeting drug-resistant bacteria, and a deeper understanding of functional genomics.
See also
However, the technology presents severe biosecurity risks. While standard LLMs have previously raised concerns by outlining steps for biological weapon creation, models capable of autonomously designing functional genomes elevate the threat vector. In unmonitored hands, future iterations could theoretically be engineered to target human hosts or bypass natural immune responses.
As genome language models advance at a rapid pace, the experiment underscores an urgent global imperative: establishing rigorous biosecurity guardrails and screening protocols before dangerous design capabilities outpace regulatory oversight.
