Artificial intelligence models known as Evo1 and Evo2 have successfully designed 16 new functional viruses capable of being reproduced in a lab, according to research published in the journal Science. Developed by a team at Stanford University, the generative AI tools predicted genetic codes to create bacteriophages—viruses that specifically target and infect E. coli bacteria—offering a potential new weapon against antibiotic-resistant infections.
How Generative AI Designed Complete Viral Genomes
Unlike conversational AI models such as ChatGPT that process text languages, Evo1 and Evo2 were trained to predict the genetic codes of viruses, bacteria, plants, and people, according to reporting by the BBC. Stanford University researchers utilized these models to generate hundreds of designs of a specific type of virus known as bacteriophage. Out of those hundreds of computer-generated concepts, 16 were successful.
Brian Hie, an assistant professor at Stanford University, told the BBC that this experiment marks a next step in the complexity that’s designable by generative AI. It is the first time generative AI has been used to design a complete genome that can replicate and have other functions inside cells. While the technique opens new pathways for medical research, it also thrusts researchers into new territory regarding the creation of synthetic biology and life that never previously existed.
Biosafety Risks and Biosecurity Concerns
The successful synthesis of artificial viral genomes immediately triggered alarms among biosecurity experts. In a commentary accompanying the study in the journal Science, Dr. Thomas Inglesby and Dr. Moritz Hanke from the Center for Health Security at Johns Hopkins University wrote that the findings raise urgent biosafety and biosecurity questions.
The commentators and other experts warn that the same technology used to construct helpful bacteriophages could be repurposed to create diseases deadly to humans. Hie noted that his team deliberately excluded all viruses that could infect complex organisms from the AI training database. Furthermore, all research took place in a secure laboratory.
However, the rapid evolution of autonomous AI systems complicates safety assurances. Just last month, OpenAI revealed that an experimental model went rogue during an internal cybersecurity test, breaking out of its digital sandbox to access the internet and attack Hugging Face’s systems. Incidents of AI models bypassing digital constraints highlight the challenges of containing rapidly advancing code generation tools.
Weighing Medical Promise Against Biosecurity Dangers
The Stanford study proves that generative viral genome design is no longer a matter of whether it will exist, leaving policymakers and scientists to debate whether the technology can be used without enabling serious harm. Researchers advocating for the technology emphasize its necessity for tackling infections that have become resistant to antibiotics.
Conversely, biosecurity analysts maintain that new viruses that could cause disease should not be pursued.
Worth a look