AI designs 16 novel viruses from DNA patterns learned across trillions of nucleotides

Researchers trained Evo, a genomic AI, on more than 9 trillion nucleotide sequences to learn DNA design rules, then used it to generate about 285 candidate viral genomes; 16 of these synthesized genomes assembled into viable bacteriophages that can infect and reproduce in E. coli, demonstrating AI’s potential to design new organisms while highlighting serious biosafety and governance concerns about applying such methods to pathogens.