New Era of Biotechnology: AI Now Capable of Designing Viruses
Scientists have successfully used artificial intelligence to design viruses that can be created and function in a laboratory, marking a new chapter for AI in biology. The breakthrough opens significant opportunities for drug development but simultaneously raises serious concerns about biological safety and the potential for misuse.
The viruses designed in the study are bacteriophages, a type of virus that specifically attacks bacteria. Bacteriophages have long been studied and used as an alternative for treating bacterial infections, including cases where bacteria have become resistant to certain treatments. In laboratory tests, a mixture of the AI-designed viruses successfully killed E. coli bacteria that were resistant to natural bacteriophages.
Dr Brian Hie, a chemical engineer at Stanford University in California, utilised an AI model designed to understand and generate genetic sequences, analogous to a ‘language model’ for genetic code. ‘The ability to rapidly design genomes and tailor them to specific microbes while overcoming resistance could transform phage therapy and expand the biotechnology toolkit,’ the researchers wrote in the journal Science.
Beyond the potential benefits, the scientists said the research raises important considerations regarding biosafety, biocontrol, and biosecurity. They urged others designing whole genomes to consult with safety and security experts throughout their projects. In an accompanying article, Prof Tom Inglesby and Dr Moritz Hanke from the Center for Health Security at Johns Hopkins University in Baltimore reinforced the warning. They noted that whilst the breakthrough is promising for life science applications, it also raises urgent questions related to biosafety and biosecurity. ‘The ability to construct viral genomes using generative AI now exists; however, the governance to steer it safely does not yet exist,’ they stated.
Hie and his colleagues used AI models named Evo1 and Evo2 to design the new viral genomes. These models were trained on genetic data from two million bacteriophages. Genetic code for viruses that can infect plants, humans, or other animals was deliberately excluded from the training data to reduce the risk of the AI designing dangerous viruses.
Dr Filippa Lentzos, a senior lecturer in science and international security at King’s College London, said the most critical point for intervention is at the DNA synthesis stage. She argued it is important to look at the broader governance picture and not focus regulation solely on AI models. ‘A layered approach makes more sense: safeguards around model development and access, responsible research review, synthesis screening, and established laboratory biosafety and biosecurity standards,’ she concluded.