Researchers at Stanford University used generative AI to design complete bacteriophage genomes and then synthesized and tested the resulting viruses against bacteria, including antibiotic-resistant strains. The study, published in Science, reported that 16 of roughly 300 AI designs proved viable and could kill E. coli. The work demonstrates an expansion of generative genomics: instead of designing partial genetic elements, the models created end-to-end functional viral genomes. Scientists and experts also raised concerns about biosafety and security oversight for AI systems that can produce replication-capable biological agents. The reported findings place the technology at the center of two urgent questions for drug development: whether AI-designed phage cocktails can improve antimicrobial options, and what regulatory frameworks are needed to manage higher-risk synthetic biology capabilities.