AI designs new bacteriophages, sparking biosecurity debate over future virus engineering

Researchers used an open-source AI model to create novel bacteriophages that target E. coli, prompting concerns about the pace of AI-driven biotech versus existing regulations.

An open-source artificial-intelligence system has been used to design bacteriophages that specifically infect Escherichia coli, a breakthrough reported in Science following a September 2025 bioRxiv preprint. While the engineered viruses are closely related to existing bacteriophages and pose no threat to plants or animals, scientists caution that future models could generate completely novel viral sequences, potentially facilitating bioweapon creation.

Current biosecurity regulations cover laboratory practices but have not been adapted to the upstream digital design of biology, leaving a regulatory lag. Experts such as Toby Webster and Doni Bloomfield advocate for a tiered data-access framework and enhanced screening of ordered genetic material, yet acknowledge limited knowledge about how training data translates to design capability. Microsoft’s recent patch to block AI-generated toxic proteins highlights the difficulty of enforcing safeguards. The discussion balances the promise of AI-derived therapies against the urgency of establishing robust governance before the technology outpaces oversight.