Researchers have successfully used AI to design bacteriophages that kill drug-resistant E. coli, sparking both medical optimism and warnings regarding biosafety and biosecurity.
Key facts
- •The AI models were trained on data from 2 million bacteriophages while excluding viruses that infect humans, animals, or plants.
- •Researchers generated thousands of potential genomes, selecting nearly 300 for laboratory synthesis.
- •Only 16 of the synthesized bacteriophages were viable, but they effectively targeted drug-resistant E. coli.
- •Experts from Johns Hopkins University stated that while the technology is promising, the governance to steer it safely does not yet exist.
- •Prof. Tom Ellis of Imperial College London noted that bacteriophages are the smallest and easiest genomes to synthesize.
Scientists have successfully created the first viruses designed by artificial intelligence, marking a milestone in synthetic biology. Using genome language models, researchers designed bacteriophages—viruses that infect bacteria—to combat drug-resistant E. coli in laboratory tests. While the breakthrough offers potential for new medical treatments, experts are calling for urgent governance to address the associated biosafety and biosecurity risks.
By the numbers
AI-Designed Bacteriophages
Dr. Brian Hie of Stanford University utilized AI models named Evo1 and Evo2, trained on genetic data from 2 million bacteriophages, to generate potential viral genomes. The researchers selected nearly 300 designs to synthesize in the lab. Although the process was inefficient, with only 16 viruses proving viable, a cocktail of these AI-designed bacteriophages successfully killed strains of E. coli that had resisted natural alternatives.
Safety and Governance Concerns
The research, published in the journal Science, explicitly excluded genetic data from viruses that infect humans, plants, or animals to mitigate risks. However, experts from the Johns Hopkins Center for Health Security warned that current governance is insufficient to manage the ability to compose viral genomes using generative AI. While some researchers argue that the threat of AI-designed pathogens is currently overblown compared to existing methods of modifying pathogens, others emphasize the need for a layered approach to regulation, including synthesis screening and responsible research oversight.
Advertisement
This article was independently rewritten by ManyPress editorial AI from reporting originally published by The Guardian Health.
