斯坦福大学研究人员利用生成式人工智能首次成功设计了完整的病毒基因组，并在实验室中合成了16种能够正常复制的新型病毒[1]。这些AI设计的病毒属于噬菌体，仅能感染细菌，对人类无害[1]。研究团队从302个AI设计方案中筛选出这16种有效的噬菌体，均能够杀死大肠杆菌[1]。
噬菌体基因组包含约5400个碱基对，相比之下最小活细胞的基因组约有50万个碱基对[1]。斯坦福大学研究人员布莱恩·海表示："这是生成式AI可设计的复杂性的下一步，这是生成式AI首次被用来设计完整基因组"[1]。这一突破有望为抗生素耐药性感染的治疗开辟新路径[1]。
然而，专家对这项技术的安全隐患提出了警告。约翰霍普金斯大学卫生安全中心的托马斯·英格尔斯比和莫里茨·汉克指出，研究引发了"紧迫的生物安全和生物安保问题"[1]。
Researchers at Stanford University have successfully used artificial intelligence to design complete viral genomes for the first time, creating 16 new viruses capable of replicating in laboratory conditions [1]. The AI-designed pathogens are bacteriophages—viruses that infect bacteria rather than humans—and represent a significant advancement in computational biology with potential applications for treating antibiotic-resistant infections [1].
The Stanford team screened 302 AI-generated designs to identify the 16 functional bacteriophages, which are capable of killing E. coli bacteria [1]. The genomes designed by the AI models Evo1 and Evo2 contain approximately 5,400 base pairs, substantially shorter than the smallest known cellular genomes, which contain roughly 500,000 base pairs [1]. Brian Hie, a member of the research team, emphasized the technical breakthrough, stating: "This is the next step in the complexity that generative AI can design, and this is the first time generative AI has been used to design complete genomes" [1].
However, the achievement has prompted serious biosecurity concerns from experts. Thomas Inglesby and Moritz Hank from Johns Hopkins University's Center for Health Security have warned that the research raises "urgent biosafety and biosecurity questions" [1], highlighting the dual-use implications of AI-driven viral design technology.