Stanford AI Creates 16 Novel Viruses, Raising Bioweapon Concerns
Researchers at Stanford University and the Arc Institute used an AI model called Evo, trained on trillions of DNA nucleotides, to design 16 new bacteriophages that successfully infected E. coli. Published in Science, the study demonstrates AI’s ability to create viable viruses from scratch. While harmless to humans, experts warn the technology could be misused to engineer dangerous pathogens, highlighting a gap between AI capabilities and biosafety governance.
Editorial responsibility
- No named human review is recorded for this page.
- Reports are grouped by semantic similarity and deterministic rules. Language models may assist titles, summaries, translation and cross-source analysis; the page itself is projected from evidence records.
- Current automated evidence projection
Cross-source coverage
Wire timeline
AI Creates 16 New Viable Viruses After Learning DNA Patterns from Trillions of Nucleotides
Researchers at Stanford University and the Arc Institute used an AI model called Evo, trained on over 9 trillion nucleotides from 128,000 genetic sequences, to design new viral genomes. Out of 285 AI-generated genomes tested, 16 successfully assembled into functional bacteriophages capable of infecting and reproducing inside E. coli bacteria. The study, published in Science, demonstrates that AI can learn the statistical patterns of DNA arrangement to create viable organisms. While the new viruses are harmless to humans and similar to the well-studied Phi X-174 virus, experts warn that such applications are advancing faster than necessary regulations and guardrails, raising concerns about potential misuse for dangerous pathogens.
AI Creates 16 New Viable Viruses After Learning DNA Patterns from Trillions of Nucleotides
Researchers at Stanford University and the Arc Institute used an AI model called Evo, trained on over 9 trillion nucleotides from 128,000 genetic sequences, to design new viral genomes. Out of 285 AI-generated genomes tested, 16 successfully assembled into functional bacteriophages capable of infecting and reproducing inside E. coli bacteria. The study, published in Science, demonstrates that AI can learn the statistical patterns of DNA arrangement to create viable organisms. While the new viruses are harmless to humans and similar to the well-studied Phi X-174 virus, experts warn that such applications are advancing faster than necessary regulations and guardrails, raising concerns about potential misuse for creating dangerous pathogens.
Scientists unveil first AI-designed virus
Scientists at Stanford University, in collaboration with the Arc Institute, reported in the journal Science that they successfully used artificial intelligence to design and build 16 new viruses not found in nature. Their AI tool, Evo, was trained on genetic sequences to recognize DNA patterns and generate recipes for novel viruses. While the breakthrough raises hopes for medical advances such as eliminating drug-resistant bacteria, it also sparks concerns about potential misuse to create dangerous pathogens. The study was constrained to simple viruses that only infect bacteria and pose no threat to humans. However, Johns Hopkins health security experts warned in a corresponding Science article that the ability to compose viral genomes using generative AI now exists, but governance to safely steer it does not, highlighting urgent biosafety and biosecurity questions.
Show 3 older updatesHide older updates
Scientists unveil first AI-designed virus
Scientists at Stanford University, in collaboration with the Arc Institute, reported in the journal Science that they successfully used artificial intelligence to design and build 16 new viruses not found in nature. Their AI tool, Evo, was trained on genetic sequences to recognize DNA patterns and generate recipes for novel viruses. While the breakthrough raises hopes for medical advances such as eliminating drug-resistant bacteria, it also sparks concerns about potential misuse to create dangerous pathogens. The viruses created are simple and can only infect bacteria, posing no threat to humans. However, experts from Johns Hopkins emphasize that the ability to compose viral genomes using generative AI now exists, but the governance to safely steer it does not, raising urgent biosafety and biosecurity questions.
Scientists unveil first AI-designed virus
Scientists at Stanford University, in collaboration with the Arc Institute, reported in the journal Science that they successfully used artificial intelligence to design and build 16 new viruses not found in nature. Their AI tool, Evo, was trained on genetic sequences to recognize patterns and write recipes for entirely new viruses. While the breakthrough raises hopes for medical advances such as eliminating drug-resistant bacteria, it also raises concerns about the potential misuse of the technology to invent dangerous pathogens. The viruses created are simple and can only infect bacteria, posing no threat to humans. However, Johns Hopkins health security experts emphasized that the ability to compose viral genomes using generative AI now exists, but the governance to safely steer it does not, highlighting urgent biosafety and biosecurity questions.
AI Model Trained on DNA Creates 16 Novel Viruses, Raising Bioweapon Concerns
Scientists at Stanford University trained an OpenAI generative model called Evo on millions of genomes to design fully functional viral genomes from scratch. After synthesizing the AI-designed DNA in the lab, 16 of the new phages successfully infected E. coli bacteria, displaying sequence patterns not found in nature. The study, published in Science, excluded human pathogen data, so the viruses cannot infect people. However, biosecurity experts warn this demonstrates AI's capability to invent dangerous bioweapons, calling for strict regulations. Critics argue creating human-impacting bioweapons would be far more difficult, and simple data restrictions could mitigate risks. The research also holds promise for developing treatments against antibiotic-resistant superbugs.