Scientists have revealed the first virus genomes produced by artificial intelligence. While custom‑designed viruses are already used in some applications — for example, viral vector therapies employ engineered viruses to deliver genetic material into cells — the new work generated entirely new genomes with substantially less human design input.
The approach
The research team trained an AI model on DNA sequences, in a manner analogous to how language models such as ChatGPT are trained on words. Using that model, they generated novel virus genomes. The resulting constructs included bacteriophages, which the report states are harmless to humans.
Significance
The development is notable for two main reasons:
- Automation: the AI reduces the need for manual design and expert intervention, enabling faster generation of genomes with less direct human effort.
- Potential breadth: although the published examples involved bacteriophages, the underlying method could, in principle, be adapted to other virus families, widening the range of possible applications.
Dual‑use and safety concerns
Researchers and ethicists emphasize the dual‑use risk: biological tools that enable beneficial therapies can also be repurposed to create harmful pathogens. The increased accessibility provided by AI could lower the technical barriers for misuse. Commentator Tom Chivers encapsulated this worry with the phrase sometimes called the “Moore’s law of mad science”: “Every 18 months, the minimum IQ to destroy the world drops by one point.”
Next steps and precautions
The findings call for further risk assessments, appropriate regulatory frameworks, and responsible disclosure practices by the scientific community. Even though the current demonstration used bacteriophages that do not infect humans, the broader trend of applying AI in biology underscores the need for updated oversight and safety measures.
Summary
AI has been used to produce complete virus genomes with reduced human input; the initial examples are bacteriophages harmless to people. Nonetheless, the approach raises significant dual‑use concerns that require coordinated scientific, regulatory and ethical responses.
