Fetching from the wire…
01
02
03
04
05
06
07
08
Public story · 2026-08-07 · high
The 16 genomes target bacteria, not people, and researchers want synthesis providers to screen buyers too.
Why now: The paper published August 6, and biosecurity researchers are citing it to push for tighter DNA-synthesis screening rules.
Evo and Evo2, generative AI models, designed 16 bacteriophage genomes that replicated and functioned in lab tests, per an August 6 paper in Science.
That's the first full viral genome AI has designed end to end and proven to work in a lab. The authors flag custom phage therapy for antibiotic-resistant infections as the near-term payoff, since all 16 genomes target bacteria, not human cells.
Biosecurity researchers are using the result differently, as a case study for tightening DNA-synthesis screening. Right now, screening checks the sequence someone orders. Researchers want it to check the identity of the orderer too, a category of screening the push itself implies providers don't do consistently now.
Sixteen working phages is the headline. The sharper point is that genome-design capability moved faster than the rule meant to gate who can order synthetic DNA. This paper proves the capability works. It doesn't require a single synthesis provider to change how they screen orders.
Published August 6, the paper gave biosecurity researchers their case study for the screening push. Watch whether synthesis companies add identity screening on their own, or wait for a law to force it after a bad order gets filled.
Each link below shares sources, entities, or timing with this story.
Shared entity: August / Shared topic / Earlier coverage
Both cover August; overlapping topics (capability, complete); earlier August coverage from 2026-08-05.
Shared entity: Science / Same source domain / Earlier coverage
Both cover Science; reported by the same outlet (science.org); earlier Science coverage from 2026-07-30.
Both cover Science; reported by the same outlet (science.org); earlier Science coverage from 2026-03-17.
Shared entity: August / Earlier coverage / Tension
Both cover August; earlier August coverage from 2026-08-06; pushes against this story (but).
Shared entity: Capability / Earlier coverage / Tension
Both cover Capability; earlier Capability coverage from 2026-08-06; pushes against this story (against).
Shared entity: August / Earlier coverage / Tension
Both cover August; earlier August coverage from 2026-08-06; pushes against this story (against).
Both cover August; earlier August coverage from 2026-08-05; pushes against this story (against).
Both cover August; earlier August coverage from 2026-08-05; pushes against this story (but).