AI Can Now Design Functional Viruses: Should We Be Worried?

Researchers at Stanford University successfully designed completely new functional bacteriophages from DNA sequences using a Genomic Language Model. This represents a significant case of implementing a complete and functional genome generated by AI.

The research team used the Evo 2 foundation model, trained on over 2 million bacteriophage genomes. To design a specific target, a ΦX174-like virus, they fine-tuned the model with 15,000 related genomes and generated 302 candidate genomes through computational constraints and quality control filters.

As a result, they successfully 'rebooted' 16 functional phages from synthesized DNA sequences, demonstrating their ability to actually infect and kill bacteria.

This technology has the following dual nature:

• Medical Innovation: It opens the possibility of customized phage therapy that precisely targets antibiotic-resistant bacteria.
• Biosecurity Risk: Concerns are raised that generative AI technology could be misused to design biological agents that cause disease.

Although it currently requires highly specialized knowledge and significant synthesis costs, the fact that AI has been proven capable of directly writing functional viral genomes suggests a technological turning point.