AI-designed viruses used to treat a patient for the first time: Radical Numerics' phages target drug-resistant E. coli

A US biotechnology startup, Radical Numerics, says viruses designed with artificial intelligence have been given to a human patient for the first time, as an experimental treatment for an E. coli infection that resisted antibiotics and existing phage treatments, ProPakistani reported, drawing on a report in The New York Times. The company announced it on Tuesday 20 October 2026.

The US Food and Drug Administration (FDA) authorised the treatment for this one patient. The company has not said whether the patient has recovered, and its findings have not been peer-reviewed or published in a journal.

How it works

The treatment is phage therapy: bacteriophages are viruses that infect and kill bacteria but not human cells.

1. The team started with T7, a well-known phage that attacks E. coli.
2. Its AI model, Omnii, which reads DNA sequences, protein structures and other biological data, was used to redesign T7's leg-like tail structures so the virus could latch on to strains of E. coli that existing phages cannot infect.
3. The scientists made the new viral DNA, put it into bacteria to produce the viruses, and showed in lab tests that they could infect and kill the target bacteria.
4. After safety checks, the viruses were handed to the patient's doctors.

The patient and the doctors have not been named.

Where it came from

The science grew out of 2025 work at the Arc Institute in California, where an AI model called Evo 2, trained on millions of DNA sequences, generated about 700,000 viral genomes. Researchers picked 285 to make in the lab and 16 produced working viruses. Radical Numerics chief executive Eric Nguyen and two other researchers from that project founded the company. Omnii was introduced in June 2026. Nguyen says the team believes personalised treatments could one day be designed within days.

Why it matters in Pakistan

Antibiotic resistance is a problem doctors in Pakistan deal with every day, from urinary and wound infections to drug-resistant typhoid. If AI can design a phage for a specific resistant strain quickly, it could one day offer a route when antibiotics fail. For now this is one patient, an undisclosed outcome and unpublished data, so it is a first step, not a cure.

What you can do now is help stop resistance spreading: take antibiotics only when a doctor prescribes them, finish the course, and buy medicines only from licensed pharmacies. DRAP regularly recalls fake batches of common antibiotics; see our report on the latest DRAP fake medicine alert.