Artificial intelligence is no longer limited to analyzing genetic code. A new study shows that AI models can design entirely new viral genomes, with researchers successfully creating functioning viruses capable of reproducing inside bacteria.

Researchers from Stanford University and the Arc Institute used AI models to generate approximately 700,000 candidate genetic sequences. Of those, 16 ultimately became functioning viruses.

The study, published in the journal Science and reported by The New York Times, was deliberately restricted to viruses that cannot infect humans, animals, plants or fungi.

Researchers used the Evo 1 and Evo 2 AI models, which were trained on trillions of nucleotides, the basic building blocks of DNA. The models were subsequently trained using approximately 15,000 viruses from a family that includes Phi X-174, a tiny virus that infects E. coli bacteria.

From the roughly 700,000 sequences generated by the AI models, researchers selected 285 candidates they considered particularly promising. Scientists then produced their DNA and introduced it into bacteria.

Sixteen of the designs successfully developed into functioning viruses capable of reproducing.

According to the study, the newly designed viruses were at least as viable as Phi X-174, while some reproduced more rapidly than the naturally occurring virus.

Researchers say the technology could eventually have beneficial applications, including advancing genetic therapies, studying genomes and developing new tools to combat harmful bacteria.

At the same time, the breakthrough raises significant questions about the potential risks posed by increasingly powerful AI systems capable of designing biological material.

The researchers emphasized that the experiment did not involve viruses capable of infecting humans and was specifically limited to bacteriophages, viruses that infect bacteria.

Still, the ability of an artificial intelligence model to generate genetic sequences that can be physically produced and become functioning, self-replicating viruses demonstrates how rapidly AI capabilities in biology are advancing.

(YWN World Headquarters – NYC)