AI Designs 16 Novel Bacteriophages, Forcing Biosecurity Rethink

Like the 1975 Asilomar conference that scrambled to set rules after recombinant-DNA breakthroughs, or the 1990s when PCR’s democratization worried forensic scientists, this milestone echoes moments when a lab technique leapt from elite craft to push-button capability (cf. genome-length horsepox synthesis, 2017). It sits at the convergence of two long arcs: the plummeting cost of DNA assembly (orders of magnitude since 2000) and the rise of foundation models that convert tacit expert rules into software. Over a century, the ability to algorithmically search genomic space could transform medicine—custom phage cocktails against the projected 10 million annual AMR deaths by 2050—yet also lowers the skill barrier for state or non-state actors to prototype novel pathogens, a risk traditionally gated by wet-lab expertise. Whether governments craft governance akin to the Nuclear Non-Proliferation Treaty (1968) or repeat the delayed regulation of cyberspace will shape whether this breakthrough is remembered as penicillin or Pandora’s box.