Startup says it used a virus created by AI to treat antibiotic-resistant infection. American startup used AI to develop personalized phages against a strain of E. coli that did not respond to antibiotics. Bacterium: American startup used artificial intelligence to develop personalized versions capable of attacking a specific strain of E. coli (Freepik) Published on October 23, 2026, at 11:44 a.m. Follow all about artificial intelligence A startup in the United States claims to have used viruses developed by artificial intelligence to treat a patient with an antibiotic-resistant bacterial infection. The procedure was announced last Tuesday, the 20th, by Radical Numerics, the company responsible for the technology. According to the technical statement published by the company, the viruses managed to eliminate, in the laboratory, bacteria taken from the patient. Subsequently, a combination of these microorganisms was administered as an experimental treatment. Radical Numerics claims that this was the first clinical application of AI-designed bacteriophages. However, the patient's results have not yet been disclosed. The report was made available on the startup's website, without publication in a scientific journal or peer review. How the virus created by artificial intelligence works. The viruses used are called bacteriophages, also known as phages. Unlike viruses that infect humans, these microorganisms attack specific bacteria. To develop new versions, Radical Numerics used an artificial intelligence model called Omnii-1, trained with genetic sequences and other biological data. The system was used to modify parts of the DNA of phages responsible for recognizing bacteria. One of the experiments described in the report involved bacteriophage T7, capable of infecting certain strains of Escherichia coli (E. coli). The researchers modified a protein located in the structure used by the virus to attach to the bacterial surface. The goal was to create variants capable of recognizing different strains without losing the ability to infect them. In laboratory tests, the company claims to have identified viable phages even after significant changes in their genetic sequences. How the viruses were used in the patient. Radical Numerics reported that it developed a combination of phages adapted to an antibiotic-resistant infection. Before administration, the researchers tested the viruses against bacteria isolated from the patient. According to the report, the combination eliminated the bacterial isolate in tests carried out in the laboratory, a procedure known as in vitro. After this stage, the phages were administered under a compassionate use regimen, with regulatory authorization and medical monitoring. This modality allows access to experimental treatments outside conventional clinical trials in certain situations, especially when patients have serious illnesses and do not have satisfactory alternatives. The company did not reveal the patient's identity, the type of infection, or the clinical results, citing the need to preserve privacy. It also did not report whether there was improvement in the condition or elimination of the bacteria after treatment. Why can bacteriophages fight resistant bacteria? Phage therapy uses viruses capable of infecting and destroying bacteria. The approach is being investigated as an alternative for cases in which antibiotics do not produce the expected results. When a phage encounters a compatible bacterium, it can attach to the surface, introduce its genetic material, and use the cell to produce new viral particles. In certain types of infection, this process ends with the destruction of the bacterium. One of the main difficulties is finding phages compatible with the strain responsible for the disease. Because the action is usually specific, a virus effective against a given bacterium may not work against another strain of the same species. Radical Numerics' proposal is to use AI to identify and develop variants targeted at specific infections, reducing the need to search for suitable viruses among large collections. What did the tests with the Omnii-1 AI show? In addition to the experimental application in the patient, the report presents results of tests carried out with the Omnii-1 model. In one of the evaluations, the researchers analyzed variants of bacteriophage T7 with alterations in proteins responsible for binding to bacteria. Among the versions with 11 to 15 amino acid changes, 69.9% were recovered in the experiments, compared with 4.8% of randomly generated controls. The company also evaluated the system's ability to predict which phages could infect different bacterial strains. The experiments included bacteria such as E. coli and Pseudomonas aeruginosa, both used in the model evaluations. These results concern the technology's performance in the laboratory and do not, by themselves, demonstrate the treatment's efficacy in humans. The company itself acknowledges that the investigated therapies have not received approval for generalized clinical use and that experience with a single patient does not allow establishing safety or efficacy.

Radical Numerics, an AI research lab building general biological intelligence, has launched Omnii-1, a biological AI model that learns across DNA, RNA, proteins, and other molecular data. Researchers used it to design a bacteriophage given to a patient with a drug-resistant infection, which the company says is the first AI-designed phage administered to a human. Omnii-1 is a genome language model that converts 28 types of biological data into a common representation. At launch, the company is demonstrating eight applications, including personalised cancer vaccine design and detection of emerging biological threats. Researchers designed the phages within weeks. A cocktail containing one killed the patient's bacterial isolate in laboratory tests before being given under compassionate use. Radical Numerics is also developing Omnii for biodefense and collaborating with Lawrence Livermore National Laboratory.