Researchers from Stanford University and the Palo Alto-based Arc Institute have used an artificial intelligence model to create viruses that do not exist in nature, marking a first-of-its-kind achievement in biological research.
The study, published in the journal Science, demonstrates how scientists successfully trained an AI model to design new viruses, potentially opening new possibilities for future research into diseases, biotechnology and medicine.
AI-Designed Viruses Created for Scientific Research
The researchers emphasized that the viruses created in the laboratory do not pose a threat to humans. However, a separate paper published in the same issue of Science warned that the technology could potentially be misused and called for urgent safety regulations.
“Although this is promising for life sciences applications, it also raises urgent biosafety and biosecurity questions,” the accompanying paper stated. Researchers noted that the ability to generate viral genomes using generative AI now exists, while governance frameworks to manage the risks are still developing.
How AI Generated New Viruses
The process behind creating the viruses follows a similar principle to how large language models such as ChatGPT, Claude and Gemini operate.
While chatbots are trained on large collections of written content to predict patterns and generate responses, the AI model used by researchers was trained on genetic data to identify patterns within viral genomes.
According to the study, the models were trained on millions of natural genomes to generate viruses designed to achieve specific biological outcomes.
The researchers used the AI system to create viruses based on Phi X 174, a naturally occurring virus that only infects E. coli bacteria. Out of nearly 300 unique viruses generated by the model, researchers identified 16 viable viruses that were able to successfully infect bacteria.
Researchers Added Safeguards to Limit Risks
The scientists implemented several safeguards to reduce potential risks. According to reports, the AI model was not provided with information about viruses capable of infecting humans or those that could threaten plants and animals.
Despite these precautions, experts have raised concerns about whether similar technologies could eventually be used to create biological threats.
Experts Warn of Future Biosecurity Challenges
Scientists not involved in the research said the technology represents an important advancement but also highlights the need for stronger oversight.
Tom Ellis, a synthetic genome engineering professor at Imperial College London, said the viruses created in the study were biologically simple and that the likelihood of an immediate catastrophe remains low.
“This is literally the smallest and easiest genome to make,” Ellis said, noting that the research involved relatively simple viral systems rather than complex human pathogens.
