Stanford and Arc Institute Scientists Use AI to Design 16 Viruses That Kill Bacteria
This digest was compiled by AI from multiple sources — links to the originals are below.
A team from Stanford University and the Arc Institute has created 16 novel viruses capable of killing bacteria, using an AI model called Evo. The peer-reviewed study, published Thursday in Science, showed the synthetic viruses were as robust as natural ones, with some replicating faster. The research exposes a regulatory gap, as U.S. rules on dangerous pathogens do not cover computational design of viral DNA.
The Evo Model
Evo was trained on about nine trillion nucleotides from millions of organisms, then fine‑tuned on the phage Phi X‑174 and its relatives. The model proposed 700,000 possible genomes, from which the team selected the most promising candidates. Doctoral student Samuel King described the step as 'the obvious next move.'
Synthesized Viruses
Out of 285 chemically synthesized DNA sequences inserted into bacteria, 16 yielded viruses that could replicate. These viruses matched natural ones in robustness, and some replicated faster than Phi X‑174. Oliver Crook of Oxford University noted they are 'not just sickly versions of stuff that already exists,' while Patrick Cai of Manchester called the work an 'important milestone.' Crook tempered expectations, adding that the AI did not invent anything fundamentally new—the viruses remain very similar to natural species.
Regulatory Gap
In late July, the U.S. National Institutes of Health issued a policy banning experiments that make pathogens more dangerous, but it exempted purely computational work unless it involves an entity of concern. Moritz Hanke of the Johns Hopkins Center for Health Security warned of a disconnect: 'What is the risk of what I've never seen before?' The study exposes a blind spot, as AI‑designed viral genomes are not clearly classified under current rules.
What's Next
The research team may next explore whether Evo can design viruses from other groups, which could yield tools for medicine and biotechnology. It remains unclear whether regulators will update biosafety rules to cover computational design of viral genomes, as experts call for closing the disconnect between research and guardrails.
1 source
Stanford and Arc Institute Scientists Use AI to Design 16 Viruses That Kill Bacteria
