Could This Harm Humans? AI Has Produced a Viable Virus Never Found in Nature
As generative artificial intelligence increasingly intersects with biology, the latest experiment highlights how quickly the field is advancing. A team at the Arc Institute in California relied on the AI model Evo to generate roughly 700,000 candidate viral genomes. From that pool, they selected about 300 for lab synthesis. Sixteen of those turned out to be viable and capable of infecting bacteria. The starting point was Phi X-174, a virus that attacks bacterial cells, so the resulting agents are limited to microorganisms and cannot harm people.
Biosecurity Challenges Come into Focus
Evo learned to read DNA patterns across many viruses, which let it propose a wide array of possible genomes. After the team narrowed the list to around 300 candidates and synthesized them, 16 proved functional. Those 16 are restricted to bacteria, yet several managed to bypass the defenses of two Escherichia coli strains, a result that has caught the attention of biosecurity specialists.
Professor Tom Inglsby and Dr. Moritz Hanke of the Johns Hopkins University Center for Health Security have pointed out that generative AI has already made viral-genome creation a reality, while safety rules have not kept pace with the technology.
The United States has introduced stricter rules for higher-risk biological research, barring experiments meant to enhance the danger posed by biological agents. However, computational work that uses AI to design new agents remains outside the ban, unless the targets are formally classified as hazardous under those regulations.
Synthesizing viruses in the lab is not a new technique, but the current findings underscore the biosecurity dilemmas facing researchers. To land on 16 working viruses, the team screened hundreds of thousands of computer-generated designs. The research's companion material was published in Science on 09.08.2026.
This experiment illustrates how quickly AI tools are moving into biology, bringing fresh questions about ethics and oversight. Although the newly created viruses are harmless to humans, their capacity to infect bacteria could carry significant implications for medicine, particularly in the context of antibiotic resistance. The scientific community must find a balance between innovation and safety to avoid future pitfalls.
As researchers delve deeper into the implications of AI-generated viruses, it's crucial to understand how these advancements can intersect with existing biological threats. A recent study highlights that dead cells may serve as covert havens for viruses, allowing them to evade the immune system. This connection underscores the complex challenges faced in biosecurity and the need for comprehensive safety measures. To explore this further, read about how viruses have adapted to hide within dead cells.