AI-generated viruses raise biosecurity concerns
Scientists have developed an artificial intelligence program capable of creating new viral genomes distinct from any known natural viruses, specifically targeting individual hosts, according to a new study, reports BritPanorama.
The study, published in the journal Science, indicates that the AI tool Evo utilized genetic sequences from millions of sources across all domains of life, paralleling how language models such as ChatGPT are trained on extensive text datasets.
Researchers from Stanford University and the Arc Institute employed the AI to generate thousands of genome combinations, particularly focusing on variants that could be hosted by E. coli bacteria cells. From these attempts, 300 genomes were synthesized and tested in the laboratory, with 16 emerging as viable viral entities.
The newly created viruses are bacteriophages, which infect bacteria but pose no harm to humans. The introduction of these AI-generated genomes marks a significant advancement, especially in the context of addressing rising drug resistance in pathogens like E. coli.
According to the research, the genetic library facilitated the AI’s understanding of natural genomes’ evolutionary constraints. Notably, one virus exhibited characteristics that were evolutionarily distant from any known viruses, indicating that the AI had initiated changes that natural evolution might take millions of years to achieve.
Tests revealed that a combination of the novel viruses could effectively combat antibacterial resistance in various E. coli strains, an outcome not achievable with naturally sourced phages. This breakthrough could pave the way for developing resilient phage therapies tailored to rapidly evolving pathogens.
“The breakthrough achieved is significant,” remarked Jordi García Ojalvo, a professor of systems biology at Pompeu Fabra University in Barcelona. This represents a crucial step in the ongoing battle against drug-resistant infections.
However, this progress is not without its reservations. Medical experts from the Johns Hopkins Center for Health Security emphasized the urgent biosafety and biosecurity questions arising from such advancements, noting that while the potential benefits for life sciences are considerable, the capacity to create viral genomes via generative AI introduces complex governance challenges.
“The ability to compose viral genomes using generative AI now exists; the governance to safely steer it does not,” they observed in a corresponding article. They have expressed concerns over the implications of pursuing eukaryote-infecting pathogens, which could result in new infectious threats to humans, animals, or plants.
While García Ojalvo assessed the risks posed by this research as comparatively modest—given that each genome must be individually tested post-design, resulting in a low yield of viable viruses—he acknowledged the complexities involved in navigating this emerging field.
The ongoing discourse surrounding these findings underscores the necessity for prudent oversight as the intersection of AI and virology continues to develop.
