Stanford’s Evo 2 AI Writes DNA to Create Virus-Killing Phages
Stanford researchers have achieved a milestone in synthetic biology by using the generative AI system Evo 2 to design entirely new bacteriophages—viruses that specifically target and kill bacteria. The team synthesized nearly 300 phages from DNA sequences produced by the model, with several demonstrating successful infection and lysis of E. coli cultures in laboratory tests. This work represents a major step forward in understanding how generative models can move beyond text and images to design functional biological systems, bridging the gap between digital sequence generation and physical, living outcomes.
What makes this accomplishment striking is its direct relevance to the growing crisis of antibiotic resistance, as these custom-designed phages could offer a precise, living alternative to traditional drugs. The process builds on a foundational understanding of What is AI in biological contexts, where models learn the grammar and syntax of genomic data to predict and create novel sequences. Furthermore, the researchers leveraged AI Tokens to represent DNA base pairs as discrete units, allowing the model to process and generate genetic code with surprising accuracy and efficiency.
Although the phages were generated by vast AI Models trained on massive genomic datasets, the selection process required significant human oversight to identify functional candidates, highlighting the current limits of full automation in biology. The open-source release of Evo 2 and its associated data will allow other labs to replicate and build upon this research, accelerating the pace of discovery in genome design and biosecurity. As the technology matures, the ability to program life at the sequence level could transform medicine, agriculture, and materials science.
- Why it matters: Offers a scalable path to create targeted therapies against drug-resistant bacteria like MRSA and E. coli, which kill millions annually.
- Why it matters: Demonstrates that generative AI can design functional biological entities, opening new frontiers for synthetic biology and genome engineering.
- Why it matters: Raises critical biosecurity questions as AI-designed biological weapons become theoretically possible, necessitating new governance frameworks.