AI designs 16 new viruses! Capable of killing drug-resistant E. coli, but raises safety concerns

In a groundbreaking development, researchers at Stanford University have utilized artificial intelligence to design novel bacteriophages, viruses that infect bacteria. The study successfully generated 16 functional viruses in a laboratory setting, with some demonstrating the ability to neutralize antibiotic-resistant strains of E. coli when used in combination.

“

Generative AI now has the capability to design viral genomes, but the corresponding safety governance has not kept up.

The AI models, Evo 1 and Evo 2, were trained on extensive bacteriophage genomic data to design complete, new genomes. Bacteriophages are being explored as a potential therapeutic tool against infections caused by antibiotic-resistant bacteria. The research team tested nearly 300 AI-designed genomes, with 16 proving to be viable and operational bacteriophages. These newly designed phages exhibited characteristics distinct from their natural counterparts and, when deployed together, rapidly overcame resistance that three strains of E. coli had developed against existing phages.

This capability holds significant promise for accelerating the development of phage therapy, particularly in combating the growing threat of antibiotic resistance. However, the research has also brought biosafety concerns to the forefront. Researchers at Johns Hopkins University's Center for Health Security noted in a concurrent commentary that while generative AI is now capable of designing viral genomes, the corresponding safety governance frameworks have not kept pace.

“

This method may accelerate the development of phage therapy in the future, especially in the face of bacteria that continuously develop resistance.

The Stanford team acknowledged these concerns and stated they deliberately excluded data on viruses that infect humans, animals, or plants to mitigate the risk of creating dangerous pathogens. Scientists also cautioned that the current technology has limitations. The study focused on relatively simple bacteriophage genomes, and it remains unproven whether AI can design more complex viruses capable of infecting humans or other animals. This research underscores the need for synchronized development of safeguards in AI, DNA synthesis, and laboratory safety management, as highlighted by researchers at Imperial College London.

“

The ability of this technology is still significantly limited.