AI-Designed Virus Given to Patient in Experimental Fight Against Resistant Bacteria
Radical Numerics says a virus designed by its AI model has been given to a patient with an antibiotic-resistant infection, marking what the San Francisco
Radical Numerics says a virus designed by its AI model has been given to a patient with an antibiotic-resistant infection, marking what the San Francisco lab describes as a first for AI-designed organisms used as medicine. Whether the treatment helped—or caused any harm—has not been disclosed.
The bacteriophages were created with the company’s Omnii-1 genome model and administered under a physician-led compassionate-use pathway authorized by the U.S. Food and Drug Administration. The lab announced the treatment in October 2026, but has not released clinical results.
Reworking a virus to target bacteria
Bacteriophages are viruses that infect bacteria rather than human cells. Researchers have explored them as a way to tackle bacterial infections, and the approach could be relevant when bacteria resist antibiotics.
Radical Numerics says its team altered the gene for the tail fiber of T7, a bacteriophage. That protein helps the virus attach to bacteria. The design aims to direct the phage toward its bacterial target; the company has not identified the patient’s infection or disclosed further details of the treatment.
The intervention remains experimental. A compassionate-use authorization allows treatment under a physician’s care, but does not establish that a therapy is effective or safe for broader use.
Evidence remains the key question
No results have been published or peer-reviewed, and the company has not revealed whether the patient improved. That leaves the central questions unanswered: whether the designed phages reached and attacked the intended bacteria, and whether the treatment was safe for the patient.
The claim, if borne out, would mark a notable step for AI-assisted biology: an AI-designed organism used directly as medicine. But a first administration is not proof of a working treatment. Evidence from the patient’s outcome, followed by research capable of assessing safety and effectiveness, would be needed to support use beyond an individual case.
For now, Radical Numerics has announced the administration, not a clinical result. Any further findings from the patient or subsequent testing will determine whether the approach moves beyond an experimental intervention.
