AI Designed 16 Never-Before-Seen Viruses to Fight Antibiotic Resistance — And Sparks Bioweapon Concerns
For the first time ever, an artificial intelligence system has engineered a collection of entirely new, previously undiscovered viruses capable of infecting and killing specific types of bacteria. This groundbreaking breakthrough unlocks transformative new opportunities to combat the growing global crisis of antibiotic resistance, but it also ignites urgent warnings about the potential misuse of this technology to build dangerous biological weapons.
For years, scientists have been able to synthesize viruses from scratch in lab settings. These man-made viruses are typically used to develop and test new antiviral drugs and vaccines, as well as expand scientific understanding of how these microorganisms function. Until now, however, the production of these viral genomes has almost exclusively relied on replicating or modifying already known pathogens or their existing variants.
In contrast, a new study from researchers at Stanford University and the Arc Institute successfully used AI to design simple, fully functional, completely unrecorded viruses. The AI generated its designs by learning patterns from genetic sequence data collected from millions of naturally occurring organisms, including animals, plants, microbes, bacteria, and viruses.
The research team focused their work on bacteriophages — viruses that only infect bacteria, which have relatively small genomes that are easy to synthesize and manipulate under controlled lab conditions. Because they exclusively target bacteria, bacteriophages hold enormous biotechnological potential as a promising alternative to antibiotics for treating drug-resistant bacterial infections.
The creation of these entirely new viruses relied on Evo 1 and Evo 2, two foundational AI models built for computational biology applications. Both algorithms were trained on millions of genomes from all domains of life, with the goal of identifying and learning complex evolutionary patterns: how genes are typically organized, which genetic sequences are conserved across lineages, and the biological constraints that allow an organism to remain functional.
For their experiment, the team used the well-studied bacteriophage Phi X-174 — which naturally infects Escherichia coli (E. coli) — as a reference. The goal was never to replicate the existing virus; instead, Phi X-174 was only used as a guide for the algorithm to generate thousands of completely new genomes, all built with a genetic architecture compatible with infecting E. coli. In other words, the AI-derived viruses retained all the core functional organization needed to recognize the bacterium, insert their DNA, replicate their genetic material, produce new viral particles, and assemble them correctly. Even with this shared functional framework, the specific DNA sequences of the new viruses differ dramatically from those found in naturally occurring bacteriophages.
16 New Functional Viruses Created Using AI
After the AI generated its designs, researchers evaluated the thousands of AI-produced genomes to select the candidates most likely to be functional, filtering based on factors including gene organization, the presence of required regulatory elements, and other criteria aligned with the known biology of Phi X-174. This process narrowed the pool down to 300 candidate genomes, which were then artificially synthesized molecule by molecule in the lab. The synthesized genomes were then inserted into E. coli to test whether they could produce working infectious viruses.
Of the 300 synthesized genomes, only 16 gave rise to fully functional bacteriophages. All 16 feature never-before-published sequences, unique gene structures, new regulatory elements, and even varying genome sizes. Their behavior also varied widely: some infected bacteria much faster than natural variants, while others had distinct replication capabilities.
The study, published this week in the journal Science, also tested the ability of AI-generated bacteriophages to combat resistant bacteria. In the experiment, researchers exposed E. coli strains that had already developed resistance to natural Phi X-174-like phages to two separate mixtures: one of AI-designed phages, and one of natural phages similar to Phi X-174. The results showed that the AI-generated viruses were able to rapidly overcome the bacteria’s resistance and establish successful infection.
According to the study authors, this finding demonstrates “a path toward artificial intelligence–generated phage therapies against rapidly evolving bacterial pathogens.”
The Two Sides of This Scientific Milestone
The discovery opens up transformative new possibilities to tackle the growing global problem of antibiotic resistance. Researchers note that this approach could speed the development of personalized treatments that evolve at nearly the same rate as the pathogens they target.
But while the milestone represents a major advance for molecular biomedicine, it also raises serious concerns about potential malicious use of the technology. Risks include engineering new human diseases, highly toxic biological substances, or pandemic-capable pathogens that could trigger a global public health crisis.
Moritz Hanke, a researcher at the Johns Hopkins Center for Health Security, argues that there are currently no safeguards capable of effectively preventing bad actors from using AI to create a lethal virus. He told The New York Times that there is a “huge disconnect” between the speed at which science and technology are advancing in this field, and the development of effective regulatory frameworks to manage risk.
Debate over these risks is not new. Three years ago, a study from the Rand Corporation warned that the most advanced AI systems available at the time already had the capacity to refine the planning and execution of biological weapon attacks. Now, with the rapid development of this technology, fears are growing that such capabilities will become even more powerful and sophisticated. The nonprofit organization also previously warned that the speed of AI evolution often outpaces governments’ capacity for regulatory oversight.
This piece was originally published on WIRED en Español and has been adapted from its original Spanish translation.
