Researchers at Stanford and the Arc Institute used large language models trained on phage genomes to generate entirely new bacteriophage genomes, reporting that 16 viable phages killed antibiotic-resistant E. coli strains. The work, published in Science, demonstrates end-to-end synthetic genome design and points to possible antimicrobial applications using phage cocktails. At the same time, the paper and associated reporting raise biosafety questions around regulatory oversight and safe governance of AI-driven biological design. The investigators’ stated intent is to improve options against drug-resistant infections, where resistance to a single phage can fail quickly—supporting the rationale for multi-phage mixtures. For developers of antimicrobial platform technologies, the dual message is operational: AI can accelerate discovery of functional biological agents, but oversight and containment frameworks will need to keep pace with capability.