Researchers at the University of Chicago used artificial intelligence to design 16 entirely new viruses from scratch, demonstrating both the creative and destructive potential of AI in biological research.
The team trained machine learning models to generate viral genetic sequences that had never existed in nature. They then synthesized these AI-designed sequences in the laboratory and confirmed that all 16 viruses could infect cells and replicate successfully. The work represents a watershed moment in synthetic biology, where computational power now enables scientists to engineer pathogens without relying on natural templates.
The researchers published their findings to advance understanding of how viruses function at the molecular level. By studying these novel constructs, scientists can better grasp viral evolution and potentially develop more robust antiviral strategies. The team emphasized that their viruses posed minimal real-world risk because they targeted only specific laboratory organisms and lacked the genetic machinery needed to infect humans or cause widespread harm.
However, the study raises serious biosecurity concerns that ethicists and policymakers cannot ignore. The same AI tools used to design harmless research viruses could theoretically be repurposed to create dangerous pathogens. David Baker, a structural biologist at the University of Washington, and other experts have called for stronger oversight of dual-use research that could enable biological weapons development.
The findings arrive as scientific institutions worldwide grapple with balancing innovation against biosafety risks. Some researchers argue that transparency and open publication accelerate beneficial discoveries. Others contend that certain methodologies should remain restricted to authorized laboratories with proper containment facilities.
The University of Chicago team did not disclose exact details that would allow others to immediately recreate their viruses, a responsible approach some institutions are beginning to adopt. Yet as AI tools become more accessible and computational costs drop, controlling who can perform such research grows increasingly difficult.
This work underscores an uncomfortable reality: powerful technologies rarely remain confined to their original