AI‑Generated Viruses: Breakthrough in Synthetic Biology Raises Safety Concerns

Sun Sep 06 2026 03:22:45 GMT+0000 (Coordinated Universal Time)

U.S. researchers used AI to create 16 functional bacteriophages, a milestone for phage therapy. Experts warn the technology could enable harmful virus design.

Scientists in the United States have successfully used generative AI models to design, synthesize, and test 16 new bacteriophages that can replicate in a laboratory. The work marks the first time whole viral genomes have been crafted entirely by artificial intelligence. While the phages show promise for treating antibiotic‑resistant bacterial infections, experts caution that the same technology could be weaponised if safeguards are not strengthened. Researchers emphasise that their project excluded any genomes capable of infecting complex organisms and was performed in a secure laboratory, but the broader implications for bio‑security remain pressing.

In a landmark for synthetic biology, U.S. researchers announced that generative AI models have successfully designed and produced 16 new bacteriophages that can replicate entirely in the laboratory. The team used two AI systems, Evo1 and Evo2, trained on genetic data from viruses, bacteria, plants, and humans to predict the language of life rather than natural language.

After selecting the most promising 302 AI‑generated genomes, scientists synthesized them and tested them against Escherichia coli. Sixteen of the phages proved effective, killing the target bacteria in petri dishes and demonstrating clear lysis spots – a breakthrough that could pave the way for treatments against antibiotic‑resistant infections.

From Computer Code to Real Viruses

While the viruses are not alive, the study represents a significant step toward designing biological functions through computer models. The phage genomes are roughly 5,400 base pairs long, far shorter than a simple living cell, yet the work shows that AI can capture evolutionary design principles well enough to produce viable viruses.

Experts from Stanford and partner institutions highlight the potential for this technology to create new drugs, enzymes, and antibodies. Prof. Marc Güell described the work as a turning point, “for the first time in history, we are beginning to design biology on a computer.”

However, the same generative models raise urgent biosafety and biosecurity questions. James Gallagher, Health and Science Correspondent, notes that the research team deliberately excluded any genomes that could infect complex organisms. In a signed commentary, Dr. Thomas Inglesby and Dr. Moritz Hanke from Johns Hopkins urge that new viruses designed to cause disease should not be pursued without stringent safeguards.

Despite the promise, scientists stress that a leap to creating living organisms would require another significant advance. The current phages are far shorter than the genome of even the simplest cells, yet the researchers are already intrigued by the prospect of extending the approach to living structures.

As synthetic biology and AI converge, the scientific community faces a dual challenge: to harness the power of genome‑writing models for human benefit while ensuring robust security measures prevent malicious exploitation.