AI-designed bacteriophages created at Arc Institute infect E. coli

Researchers at the Arc Institute in Palo Alto, California, used an artificial intelligence model named Evo to design 16 new bacteriophages capable of infecting E. coli. The study, published in the journal Science on 6 August 2026, trained the AI on DNA sequences from naturally occurring viruses before tasking it to generate novel viral genomes. The team synthesised 300 of the 700,000 candidate genomes produced by Evo, and 16 proved viable in laboratory tests.

The viruses are bacteriophages, which infect bacteria but cannot infect humans. The study excluded datasets containing human pathogens from the AI’s training, ensuring the designed viruses pose no direct threat to people. The 16 viable phages overcame resistance in two strains of E. coli, demonstrating functionality beyond simple imitation of natural viruses. The research was conducted by a team including scientists from Stanford University.

An accompanying article in Science, authored by biosecurity experts at Johns Hopkins University, warned that the ability to design viral genomes using AI raises urgent biosafety and biosecurity questions. The National Institutes of Health issued new policies last week to prevent high-risk life science research but did not restrict AI-based research unless it involves an existing entity of concern, such as smallpox.

What the coverage left out

None of the right-rated digests mentioned the potential medical applications of the research, such as combating antibiotic-resistant bacteria. Neither of the right-rated digests carried the biosecurity warnings from Johns Hopkins experts or the fact that the AI generated 700,000 candidate genomes before producing 16 viable viruses. The left-rated digests did not mention the potential for medical breakthroughs.