The needle-free jet injector pressed against a volunteer’s arm in a Southampton clinic. No syringe. No jab.
Just a high-pressure stream of fluid carrying an immune-training molecule designed by a machine. That moment, part of a Phase 1 trial, is now the first published evidence that an AI-created vaccine can work in people.
The results, published in the Journal of Infection, show the vaccine is safe. More than that: it triggered immune responses not just against SARS-CoV-2, but against the original SARS virus that killed 774 people in 2003, and against bat coronaviruses that have never infected a human — but could. The researchers at the University of Cambridge and their spinout company DIOSynVax built the vaccine’s active ingredient from scratch using machine-learning algorithms.
Those algorithms combed through genetic data from every known Sarbeco coronavirus. They found the shared molecular features.
They stitched them together into a single super-antigen. This is not a tweak to an existing vaccine. It is a different logic entirely.
Traditional vaccines train the immune system to recognize a specific spike protein on a specific variant. When the virus mutates, the vaccine loses potency. That game of whack-a-mole has defined the entire COVID-19 pandemic.
The AI-designed antigen, by contrast, targets the stable, conserved parts of the virus family — the parts that do not change. The immune system learns to recognize the structural core of an entire family of threats.
What is at stake is the difference between reacting and preparing. The current vaccine paradigm is reactive. A new virus emerges.
Scientists sequence it. They design a vaccine.
They manufacture it. They distribute it. That process takes months, and during those months people die.
The Cambridge approach flips the sequence. The AI can analyze viral families in advance, design antigens for threats that do not yet exist, and have a vaccine ready before the outbreak begins. The same platform, the researchers say, could be adapted for Ebola, for influenza, for any rapidly mutating virus.
The trial enrolled healthy volunteers in Southampton and Cambridge. No significant side effects were reported.
That is the baseline for a Phase 1 trial — safety first. But the immune data is the real signal. Participants developed antibodies that neutralized multiple coronaviruses.
Not just the one that caused the pandemic, but the one that caused the 2003 SARS outbreak, and bat viruses flagged by virologists as future spillover risks. The machine found the common ground among them.
This is early. Phase 1 trials are small. They test safety and immune response, not real-world protection.
Larger studies are needed to confirm that the immune response translates into actual prevention of infection or disease. That is the hard road ahead. But the fact that an AI-designed antigen passed the first human hurdle at all is a data point that did not exist before June 5.
The needle-free delivery system matters too. It removes the fear of needles, which keeps some people from getting vaccinated.
It also eliminates the logistical burden of syringe supply and sharps disposal. A jet injector is simpler. Faster.
Cheaper to deploy at scale. The stakes are concrete.
The next pandemic virus is already circulating in some animal reservoir — a bat, a bird, a pig. It is mutating right now. The question is whether the vaccine development system will be ready before it jumps to humans.
The Cambridge trial suggests a path: use AI to map the viral family in advance, design a universal antigen, and have a safe vaccine candidate ready to test before the first human case is ever reported. That is the difference between a vaccine that arrives after the peak of a pandemic and one that is waiting when the pandemic starts.
