NewsAI

AI-Designed Viruses Become Reality: Hopes for New Medicines Meet Biosecurity Warnings

The Guardian reported on August 6 that scientists have made the first viruses designed by artificial intelligence, achieving a milestone that raises hopes for new medicines while prompting urgent biosecurity concerns. The work was published in the journal Science.

Key Takeaways

  • Scientists made the first AI-designed viruses that work, successfully killing antibiotic-resistant E coli in lab tests
  • Stanford chemical engineer Brian Hie used genome language models Evo1 and Evo2 to design the viral genomes
  • Published in Science, the study drew warnings that the ability to compose viral genomes with generative AI now exists while governance has not caught up

Designing Phages with Genome Language Models

The viruses in question are bacteriophages, which infect only bacteria and are used around the world to treat patients with persistent infections. Brian Hie, a chemical engineer at Stanford University, used genome language models, the genetic equivalent of the large language models behind AI chatbots, to design functioning bacteriophage genomes. The models were trained on genetic data from 2 million bacteriophages. Genetic code for viruses that can infect plants, humans or other animals was intentionally excluded from training to reduce the risk of designing dangerous viruses.

The AI generated thousands of candidate genomes, from which the researchers selected nearly 300 to make in the lab. Only 16 bacteriophages proved viable, but a cocktail of them swiftly overcame resistance in two strains of E coli. The researchers wrote that the ability to rapidly design genomes and tune them for specific bugs while overcoming resistance could transform phage therapy and expand biotechnological toolkits.

Biosecurity Warnings

In an accompanying article, Prof Tom Inglesby and Dr Moritz Hanke at the Center for Health Security at Johns Hopkins University reinforced the concern: “The ability to compose viral genomes using generative AI now exists; the governance to safely steer it does not.” They warned such genomes might encode new pathogens that cannot be contained by existing countermeasures, and said work on pathogens that could infect humans, animals or plants should not be pursued.

Tom Ellis, a professor of synthetic genome engineering at Imperial College London, called the work impressive but noted how hard it would be to make more complex genomes: “This is literally the smallest and easiest genome to make.” He said restricting access to genetic data and limiting the creation of genomes that look dangerous would help, adding that governments are already working on this.

TechApple’s Take: AI Goes from Writing Text to Writing Genes

The result shows how fast AI for science is moving: from generating text to designing working biological genomes in a matter of years. Notably, the team deliberately excluded dangerous virus data from training, showing the designers understood the risk. The open question is how that line holds once open-source models and datasets spread.

For the tech industry, this case is a reminder that AI safety and biosecurity must be considered together, and regulation needs to keep pace with capability. In the near term, the most direct application remains phage therapy, offering a new route against antibiotic-resistant infections, but the longer-term governance framework needs to be built quickly.

Cover image generated by AI for illustrative purposes only.

Source: The Guardian