An AI-designed virus has reportedly been given to a human patient for the first time, and the company behind it is not saying whether it worked. San Francisco lab Radical Numerics announced in October that bacteriophages designed by its new genome model, Omnii-1, were administered to a person with an antibiotic-resistant infection, according to TechRadar.

If the claim holds up, it would be the first time an AI-designed organism of any kind has been put into a person as medicine. It is also a story with a giant asterisk, because nothing about the outcome has been published, peer-reviewed or disclosed.

What phages are and why they are back

Bacteriophages are viruses that infect bacteria and ignore human cells entirely. They have been used as therapy for more than a century, predating penicillin, and they have enjoyed a quiet revival as antibiotics fail. The World Health Organization considers phage therapy broadly safe, and a phage can target one bacterial strain while leaving the rest of the microbiome alone.

That precision is also the headache. Each phage recognizes a narrow set of surface receptors on bacteria, so doctors have to find one that matches a patient's exact infection. Clinicians search existing phage libraries, then environmental samples, and only try engineering as a last resort. For someone with sepsis, that search is a race against time.

How Omnii-1 edits a virus

The scale of the underlying problem is not in dispute. The Global Burden of Disease project estimated that antimicrobial resistance was associated with roughly four point seven million deaths in twenty twenty-one, with close to forty million deaths projected between twenty twenty-five and twenty fifty.

Radical Numerics was founded by members of the team behind Evo and Evo two, the Arc Institute models that produced the first entirely AI-written phage genomes. That earlier work generated seven hundred thousand candidate genomes, of which two hundred eighty-five were synthesized and sixteen produced viable viruses.

Omnii-1 takes a more surgical approach. Instead of inventing whole genomes, it edits the targeting part of a known virus. The team worked with T7, a well-studied E. coli phage, and rewrote the gene for its receptor-binding tail fiber. The company reports that sixty-nine point nine percent of its designed variants were recovered as viable phages, versus four point eight percent for random controls.

The patient and the silence

The treatment was given under a physician-led compassionate-use pathway authorized by the FDA. The patient reportedly had an E. coli infection that had resisted antibiotics and every available phage. The company tested its cocktail against bacteria isolated from that patient and confirmed it killed the isolate in the lab before administering it.

After that, silence. Radical Numerics declined to describe the clinical course, outcome, hospital or physicians, citing patient privacy. As of October twenty-ninth, trade coverage for infectious disease clinicians still reported no disclosed outcome and no peer-reviewed paper, Contagion Live noted. The company's own technical write-up says single-patient observations cannot establish safety or efficacy, and its CEO has said it intends to publish through peer review.

Outside experts sit between intrigued and unconvinced. Harvard microbiologist Michael Baym, who was not involved, said the result would be incredible if it works, but he is skeptical until the treatment finishes and the viruses can be examined.

Promise, profit and the dual-use problem

The commercial logic is easy to see. Phage hunting is slow, and Radical Numerics believes it can compress design from months to days. Ottawa Hospital infectious disease clinician Marisa Azad, who is collaborating with the company, has described the conventional search as remarkably time-consuming. The treatment is proprietary, and the company hopes to commercialize it.

There is a darker side too. A model that can design viruses to order is, structurally, a model that can design viruses to order. Radical Numerics says it is committing to biosafety, containment and data-sharing measures scaled to the model's capability, the same dual-use tension that follows every powerful biological design tool.

Engineered phages have reached patients before, including a twenty nineteen case that treated a drug-resistant Mycobacterium infection. What is new is the design process, not the concept, so the governance questions attach to the AI-driven method.

For now, a bold experiment was run on a very sick person, and nobody outside the company knows how it ended. Keep an eye on GenZ NewZ health and our tech coverage for what comes when the data finally lands.