AI Designs Functional Viruses, Offering Hope Against Antibiotic Resistance – and Raising Biosecurity Concerns
A team at Stanford University has achieved a groundbreaking feat: designing a fully functional virus from scratch using artificial intelligence. The research, currently available as a preprint on bioRxiv, demonstrates AI’s potential to not just mimic nature, but to improve upon it, opening exciting possibilities in fields like medicine and agriculture while simultaneously raising critical biosecurity questions.
The team focused on bacteriophages – viruses that specifically target and kill bacteria – due to their inherent safety for humans and growing relevance in combating antibiotic resistance. Researchers utilized an AI model called Evo 2 to generate the genetic code for thes phages. The results were remarkable: the AI successfully created 16 viable phages, some of which proved more effective at eliminating bacteria than naturally occurring phages. This signifies a leap beyond simple imitation; AI facilitated the creation of a superior virus.
This discovery arrives at a crucial time. The escalating crisis of antibiotic resistance threatens to overwhelm healthcare systems globally. Estimates suggest that without intervention, resistant infections could cause over 39 million deaths by 2050. Phage therapy – employing cocktails of phages to combat bacterial infections – is gaining traction as a potential solution, with treatments already available in some countries. AI-driven design could accelerate this field, enabling the rapid growth of personalized phages tailored to individual infections or patients.
Beyond human health, the applications extend to agriculture. Phages could offer a biological choice to pesticides,protecting crops from destructive bacteria and reducing food loss. Evo 2 and similar tools promise the ability to engineer “tailor-made” phages,optimized for both efficacy and safety.
However, this powerful technology isn’t without its risks. A recent article in The Washington Post highlighted concerns about the world’s unpreparedness for AI’s capacity to generate functional viruses. The core question raised is the potential for malicious actors to exploit these methods for creating risky pathogens.
The Stanford team implemented rigorous safety protocols, utilizing a harmless model and openly detailing their procedures.Nevertheless, as AI models become more refined and accessible, the possibility of misuse increases.
Dr. Hie, a member of the research team, acknowledges the concerns but remains optimistic. He emphasizes that DNA synthesis and manipulation remain complex processes, not easily accessible to everyone.Moreover, he argues that the potential benefits – potentially saving millions of lives by addressing antimicrobial resistance - substantially outweigh the risks.
The Stanford experiment represents a pivotal moment. It marks the first instance of AI successfully designing a virus from inception to completion. While current applications are limited to benign bacteriophages, the convergence of biology and AI holds the potential to fundamentally reshape medicine, agriculture, and our understanding of life itself. The challenge now lies in responsibly shaping this technology’s future, ensuring its benefits are realized while mitigating potential harms.
The preprint of the research is available on bioRxiv: https://www.biorxiv.org/content/10.1101/2025.09.12.675911v1.full
