Written By | Torch Media Africa
For the first time, a vaccine whose key active ingredient was designed entirely by artificial intelligence has successfully completed a Phase I human clinical trial
Search
CAMBRIDGE, England
Every pandemic begins the same way: a virus no one has a vaccine for, spreading through a population with no immunity. COVID-19 killed more than seven million people while scientists scrambled to catch up. The question haunting virologists is not how to respond to the last pandemic — but how to be ready for the next one before it arrives. This week, scientists at the University of Cambridge and spin-out company DIOSynVax announced a potential answer — and it may be the most consequential medical development of 2026.
For the first time, a vaccine whose key active ingredient was designed entirely by artificial intelligence has successfully completed a Phase I human clinical trial. Published June 5 in the Journal of Infection, the results describe a vaccine called pEVAC-PS — found safe, well-tolerated, and capable of triggering immune responses not just against known coronaviruses, but against bat viruses that have never infected a human.
In short: AI may have just designed a shield against pandemics that have not happened yet.
The Problem With Every Vaccine Ever Made
Almost every vaccine ever created targets a specific pathogen. The flu vaccine must be reformulated annually as the virus mutates. COVID-19 boosters chased new variants — Delta, Omicron, and beyond.
The process is reactive. Scientists cannot make a vaccine against a pathogen they don't know exists.
The Cambridge team asked a different question: what if you could design a vaccine against an entire family of viruses? Find the parts that never change — the conserved features that persist across evolution — and train the immune system to attack those.
How AI Designed A Vaccine No Human Could Have Made Alone
The researchers focused on Sarbecovirus — the coronavirus family that includes SARS-CoV-2, the original 2003 SARS virus, and dozens of bat coronaviruses considered potential future pandemic threats. The genetic diversity is immense. No human could manually design an antigen that works against all of them. But a machine could.
The team fed global genetic sequence data from the entire Sarbecovirus family into a machine learning system. The AI was asked to find the parts that don't change — structural features so fundamental that any mutation would cripple the virus itself.
From these conserved features, the AI designed a synthetic super-antigen — an artificial protein structure that captures key antigenic signatures shared across the whole family. That super-antigen is the core of pEVAC-PS. When injected, the immune system learns to recognize and attack the conserved features that all Sarbecoviruses share.
The vaccine is delivered via a needle-free jet injection — a small device that pushes the DNA-based formulation through the skin using a high-pressure stream. No needle required.
The Trial Results: Safe, Effective, And Strikingly Broad
The Phase I trial involved 39 healthy volunteers aged 18 to 50, recruited at NIHR facilities in Southampton and Cambridge.
The headline result: the vaccine was safe with no significant side effects. But the immunological results were remarkable.
The vaccine generated immune responses against:
Because pEVAC-PS targets structural features fundamental to how Sarbecoviruses function, any mutation that helps the virus escape would likely cripple the virus itself. The vaccine is, in theory, self-reinforcing against evolution.
DIOSynVax And The New Frontier Of AI-Driven Drug Discovery
DIOSynVax was founded in 2017 — before AI-driven drug discovery became fashionable. The core thesis — that machine learning could design better vaccines by processing more data, identifying subtler patterns, and exploring more of the antigen design space — was a bold bet.
Nine years later, that bet has produced the world's first successful human trial of a fully AI-designed vaccine active component.
The team is now preparing for Phase II trials, which will assess whether pEVAC-PS can generate protective immunity at a level that prevents disease. A successful Phase II would lead to Phase III — large-scale efficacy trials. A deployable universal coronavirus vaccine could be available within three to five years.
A Platform, Not Just A Product
The team describes pEVAC-PS as proof of concept for an entirely new platform. The same methodology — AI scanning global genomic databases for conserved features, designing synthetic super-antigens, delivering them via needle-free DNA platforms — could be applied to influenza, Ebola, and other virus families.
The vision: a world where governments maintain a library of broad-spectrum vaccines against dangerous virus families, ready for mass production at the first signal of an emerging outbreak.
Expert Perspective: A Paradigm Shift
Infectious disease specialists have reacted with genuine excitement. The combination of AI-driven antigen design with broadly protective vaccines is seen as potentially transformative.
Phase I trials are a first step, not a finish line. Most vaccine candidates that pass Phase I do not ultimately achieve approval. But the breadth of the immune response observed in pEVAC-PS — targeting bat coronaviruses no conventional vaccine has ever addressed — represents something genuinely new.
The needle-free delivery system is also attracting attention, as needle-based campaigns face logistical and cultural barriers in the world's most vulnerable regions.
What This Means For You
In the immediate term, pEVAC-PS will not be available to the public. Phase II and Phase III trials must be completed, and regulators must review the data.
But in the medium term, the implications are profound. If this platform fulfills its promise, the next coronavirus pandemic could meet a response measured in weeks or months rather than years. A pre-immune world would face a very different outbreak curve than 2020.
More broadly, this is a signal that AI is beginning to fulfill its most optimistic promise in biomedicine: not as a chatbot or image generator, but as a scientific collaborator capable of seeing patterns beyond human capacity — and turning those patterns into tools that save lives.
The Bottom Line
Every vaccine in history was designed by human scientists studying known pathogens. pEVAC-PS represents something different: a vaccine designed by a machine, targeting viruses that have not yet made the species jump, built on an AI's ability to see the conserved skeleton of an entire viral family.
It passed its first human test. The road ahead is long, but the destination is visible: a world where the next pandemic is met not with fear and frantic scramble, but with a vaccine already waiting on the shelf — designed, in advance, by an intelligence that never sleeps.
