Scientists use AI to create viruses not found in nature, simultaneously raising alarms, enthusiasm

Stanford researchers used the AI model EVO 2 to design 16 new bacteriophages, viruses that target bacteria, which could help fight antibiotic-resistant infections. The team synthesized and tested nearly 300 genomes, finding 16 viable ones. While the open-source tool raises biosecurity concerns, safeguards limit its ability to target animals, plants, or fungi. Experts call for updated oversight.

Researchers at Stanford University used artificial intelligence to create 16 functional viruses not found in nature. The viruses are bacteriophages, which exclusively target bacteria, and may offer a solution to the rising problem of antibiotic-resistant infections. Chemical engineer Brian Hie and Samuel King used the generative AI model EVO 2, which Hie created, to design new genomes based on the bacteriophage ΦX174, which targets E. coli. The team synthesized and tested nearly 300 genomes, finding 16 that were viable and effective. A mixture of the designed phages was able to overcome ΦX174-resistant E. coli strains. The researchers addressed biosafety and biosecurity, limiting the AI's training data so it cannot generate viruses targeting animals, plants, or fungi. However, scientists from Johns Hopkins University, Thomas Inglesby and Moritz Hanke, warned that the limitations could be circumvented and called for national and global oversight. They noted that current laws and international policies do not cover AI-generated pathogens.