AI designs new viruses infecting bacteria, 16 prove viable

Researchers have used an AI model to design entirely new viruses that can infect bacteria, according to a study published in Science on August 6, 2026. The work, carried out by scientists at Palo Alto's Arc Institute and Stanford University, produced 16 functional viruses that were at least as viable as the natural virus Phi X-174, with some reproducing faster.

The team employed genome language models called Evo 1 and Evo 2, which were trained on trillions of nucleotides to learn the "grammar" of genetic code. For this experiment, the models were further trained on about 15,000 viruses from the same family as Phi X-174, a virus that only infects E. coli bacteria, and the scope was explicitly limited to exclude anything that could infect humans, other animals, plants, or fungi.

Evo generated around 700,000 possible new viral sequences, from which the researchers selected the 285 most promising candidates. They manufactured the DNA for these candidates and inserted it into bacteria, resulting in 16 working viruses. The study highlights both the potential benefits and risks of such technology, as AI-designed viruses could improve gene therapies, help researchers better understand genomes, and produce tools for attacking harmful bacteria.

However, the same technology could be misused to design highly contagious or deadly viruses, raising concerns about the current regulatory environment. Researchers have already shown that standard chatbots can provide concerning help with planning biological weapons, and a model that can design genomes takes that threat to another level.

The scientists took appropriate precautions by limiting their work to bacteria-infecting viruses, but the study underscores the need for guardrails to keep pace with rapidly advancing AI capabilities in biology.