NCCapital

Researchers from Stanford University and the Arc Institute have developed an artificial intelligence system capable of designing entirely new viruses that do not exist in nature. The breakthrough, detailed in a study published in the journal Science, marks a significant advance in AI-assisted biology while also raising important questions about safety and oversight.

The AI model, known as Evo, was trained to understand genetic sequences and generate entirely new viral genomes. Instead of copying existing viruses, the system creates novel genetic designs based on patterns it has learned from large collections of biological data.

To reduce potential risks, the researchers deliberately excluded viruses that infect humans from the model's training data. Instead, Evo was trained only on bacteriophages—viruses that infect bacteria rather than people or animals. This approach was intended to demonstrate the technology's capabilities while minimizing the possibility of generating harmful human pathogens.

Scientists say the technology could dramatically accelerate research in medicine, biotechnology, and synthetic biology. Custom-designed viruses may eventually be used to target antibiotic-resistant bacteria, deliver gene therapies more efficiently, or create new treatments for diseases that are difficult to address using conventional methods.

Bacteriophages have attracted growing attention as a potential alternative to antibiotics. Because these viruses specifically attack bacteria, researchers believe AI-designed phages could someday be engineered to combat dangerous infections while leaving healthy human cells unaffected.

Beyond medicine, the Evo model could also help researchers better understand how viruses evolve over time. By generating new genetic sequences in a controlled environment, scientists can study which genetic changes affect viral behavior, stability, and function without relying solely on naturally occurring organisms.

Despite its scientific promise, the research has also sparked concern among biosecurity experts. Advances in AI-driven genetic design raise the possibility that similar technology could eventually be misused to create dangerous pathogens if placed in the wrong hands or developed without appropriate safeguards.

The research team acknowledged these concerns and emphasized that safety played a central role throughout the project. They intentionally limited the model's training data, evaluated potential risks before publication, and argued that responsible development and transparency are essential as AI becomes increasingly capable in biological research.

Experts say the study highlights both the opportunities and challenges created by the convergence of artificial intelligence and biotechnology. As AI systems become more sophisticated, governments, research institutions, and technology companies will likely face increasing pressure to establish standards that encourage innovation while preventing misuse.

The publication of the study in Science is expected to fuel ongoing discussions about how powerful AI tools should be governed. While Evo demonstrates the potential to accelerate medical discoveries and scientific understanding, it also serves as a reminder that advances in artificial intelligence often require equally significant investments in ethics, security, and responsible oversight.