Scientists Just Used AI to Design a Universal Coronavirus Vaccine — And It Passed Its First Human Trial

Introduction: The Vaccine That AI Wrote From Scratch

Every COVID-19 vaccine you have ever received was designed by humans. Scientists studied the virus, identified the spike protein, engineered a version of it that the immune system could learn to recognize, and manufactured it at scale. That process was faster than any vaccine development in history — and it was still, at its core, the same human-driven approach that produced every vaccine of the past century.

On June 5, 2026, that changed. Researchers at the University of Cambridge and its spinout company DIOSynVax announced that a vaccine designed entirely by artificial intelligence — not assisted or analyzed by AI, but architected from scratch by it — had successfully passed its first human clinical trial. The vaccine, called pEVAC-PS, was found to be safe, well tolerated, and capable of generating immune responses against multiple coronavirus strains simultaneously, including strains that the trial volunteers had never been exposed to and strains that have not yet crossed from animals to humans. The results were published in the Journal of Infection, one of the leading peer-reviewed journals in infectious disease research.

This is not a press release claim or a laboratory result. It is a Phase I human trial outcome. And what it represents for the future of vaccines, pandemic preparedness, and AI's role in medicine is genuinely significant.

What Is a Universal Coronavirus Vaccine and Why Has One Never Existed Before

To understand why this result matters, it helps to understand what every existing coronavirus vaccine cannot do.

SARS-CoV-2 — the virus responsible for COVID-19 — belongs to a large family of coronaviruses called Sarbecoviruses. This family includes SARS-CoV-1, which caused the 2002–2003 SARS outbreak, as well as dozens of coronavirus strains circulating in bat populations around the world. Scientists have long known that the next pandemic-scale coronavirus is likely to emerge from within that same family — the evolutionary lineage that has already produced two global health crises in twenty years.

Every COVID vaccine deployed since 2020 was designed to target the spike protein of a specific viral strain. When new variants emerge with mutations in that spike protein, the vaccine's effectiveness drops. This is why annual booster shots became a feature of COVID management — the virus keeps evolving, and the vaccines are perpetually catching up to it. The spike protein changes. The vaccine targeting the old version becomes less effective. A new shot is manufactured, tested, and distributed. The cycle repeats. This approach works well enough for managing known variants, but it does nothing to prepare the immune system for a coronavirus strain that does not yet exist.

A universal coronavirus vaccine takes a fundamentally different approach. Instead of targeting the spike protein — which mutates frequently — it targets the conserved regions of the virus: the genetic sequences and structural features that remain stable across many different coronavirus strains because they are essential to the virus's function. If those conserved features change, the virus stops working. Which means a vaccine targeting them remains effective even when the virus evolves in other ways. Designing such a target is the hard part. The human immune system needs to be pointed at something specific. Identifying which features are truly conserved, truly accessible to antibodies, and truly capable of triggering a strong immune response — across dozens of different coronavirus strains simultaneously — is a computational challenge of enormous complexity.

That is exactly the problem the AI system at DIOSynVax was built to solve.

How the AI Designed the Vaccine

The DIOSynVax AI platform began with a dataset that no individual human researcher could process: all the available genetic sequence data logged by coronavirus surveillance programs around the world, covering the entire Sarbecovirus family. Every sequenced strain. Every documented mutation. Every genetic variation recorded across decades of viral monitoring.

The AI analyzed that dataset to identify antigens — the molecular targets that a vaccine instructs the immune system to recognize and attack — with three specific properties. First, high conservation across strains: the antigen had to be present and largely unchanged across as many different coronaviruses as possible, so that immunity built against it would work broadly rather than narrowly. Second, structural accessibility: the antigen had to be physically exposed on the surface of the virus in a way that antibodies could actually reach and bind to it. Third, immunological potency: the antigen had to be capable of triggering a robust immune response rather than one the immune system would ignore or underreact to.

From that analysis, the AI designed what the researchers call a "super-antigen" — a synthetic molecular structure containing the conserved features common to the entire Sarbecovirus family, engineered to maximize the immune response it triggers. No human researcher assembled this antigen from known components. The AI generated it computationally by evaluating the entire known sequence space of the virus family and selecting a design that optimized for all three properties simultaneously. The resulting vaccine candidate — pEVAC-PS — is a DNA plasmid-based vaccine, meaning it delivers genetic instructions to cells rather than proteins or live virus particles. It was designed to be administered needle-free, through a microfluid jet device that delivers the vaccine through the skin without a syringe.

What the Trial Found — and What It Did Not Find

The Phase I clinical trial of pEVAC-PS was conducted at National Institute for Health and Care Research clinical research facilities in Southampton and Cambridge, sponsored by University Hospital Southampton NHS Foundation Trust and funded primarily by Innovate UK. It enrolled 39 healthy volunteers between the ages of 18 and 50 who had previously received two or three doses of existing COVID-19 vaccines.

Participants were sequentially enrolled across four dose-escalation cohorts, receiving doses of 0.2 mg, 0.4 mg, 0.8 mg, or 1.2 mg at two time points: day zero and day 28. The trial's primary purpose was to answer three questions: Is the vaccine safe? Does it cause unacceptable side effects? Does it generate the type of immune response it was designed to produce?

The answers were unambiguously positive on all three counts. No serious adverse events were reported across any of the four dosing cohorts. The vaccine was described as well tolerated, with no significant side effects identified at any dose level. And the vaccine was immunogenic — it generated antibody responses against SARS-CoV-2, the original SARS virus from 2002, and related bat coronaviruses that the volunteers had never been exposed to and which have not yet caused human infections.

That last finding is the one that sets this trial apart from anything that has come before it. The volunteers developed immune responses against viruses they had never encountered, targeting conserved features that those viruses share with the ones the participants were vaccinated against. If that breadth of protection holds through later-stage trials, it means the vaccine could potentially provide protection against the next coronavirus to emerge from animal populations — before it emerges, before scientists have sequenced it, before any variant-specific vaccine has been designed. That is the definition of pandemic preparedness that the world did not have before COVID-19 arrived.

What Still Has to Happen Before This Changes Medicine

A Phase I trial success is a critical milestone. It is also a long way from a licensed vaccine available at your doctor's office.

Phase I trials are designed to establish safety and tolerability at different dose levels in a small number of healthy volunteers. Thirty-nine people is enough to catch obvious safety signals — serious side effects, dangerous immune reactions, dosing problems — but it is not enough to establish how well the vaccine actually prevents infection in the real world, or how durable the immune protection it generates will be over time. Those questions require Phase II and Phase III trials, involving hundreds or thousands of participants, conducted across a broader age range including older adults and people with underlying health conditions, over timelines of months to years.

The researchers behind pEVAC-PS have not yet published data on the durability of the immune response — how long the antibodies generated by the vaccine persist at protective levels. They have not yet tested the vaccine in people over 50, the age group most vulnerable to severe coronavirus disease. And they have not yet demonstrated, in a controlled trial, that vaccination with pEVAC-PS actually prevents infection with a novel coronavirus strain. Generating immune responses and preventing infection are related but distinct outcomes, and only a large efficacy trial can establish the second from the first.

None of these limitations diminish what the Phase I result demonstrates. They are the standard, appropriate pathway that every vaccine must follow, and pEVAC-PS has now cleared the first gate. The University of Cambridge team and DIOSynVax have publicly stated their intention to proceed to further clinical development, which means the Phase II planning is already underway.

What This Means for the Future of AI in Medicine

The significance of pEVAC-PS extends beyond coronavirus vaccines. This trial is the first confirmation, in a human clinical trial, that a vaccine designed entirely by AI — not with AI assistance, not with AI-supported analysis, but with AI as the primary design engine — can produce a safe and immunologically active result in human beings.

That confirmation changes the landscape of what AI-driven drug discovery can credibly claim. Insilico Medicine's ISM001-055, the first AI-discovered drug to show positive results in Phase IIa clinical trials, demonstrated AI's capacity to identify novel drug targets. AlphaFold, which earned a Nobel Prize in Chemistry for predicting protein structures, demonstrated AI's capacity to accelerate fundamental biology research. The pEVAC-PS trial adds a third data point: AI can design the active ingredient of a vaccine from first principles, and that ingredient can work in the human body.

The practical implication is that the timeline from pathogen identification to vaccine candidate could shrink dramatically. During COVID-19, scientists identified the SARS-CoV-2 spike protein as the target within weeks and began vaccine development almost immediately — but that speed was still dependent on human researchers making design choices that took time. An AI system that can evaluate the entire genetic sequence space of a viral family and generate an optimized antigen computationally could, in principle, produce a vaccine candidate for a newly emerged pathogen in hours rather than weeks. Whether that candidate would work is still something only a clinical trial can establish. But having a candidate to test is the prerequisite for every trial.

Conclusion: The Trial That Proved a New Kind of Medicine Is Possible

Thirty-nine volunteers walked into clinical research facilities in Southampton and Cambridge this year and received an injection — delivered without a needle — of a vaccine that no human scientist had designed. The antigen inside it was generated by an AI system that read the genetic code of dozens of coronaviruses, identified what they all had in common, and built a molecular target designed to make the immune system recognize all of them at once.

The vaccine was safe. It generated immune responses. It provoked antibodies against viruses those volunteers had never encountered.

That result does not mean annual COVID boosters are canceled, or that the next pandemic is already defeated, or that AI has solved vaccine development as a category of problem. It means that the first gate is cleared, the proof of concept is established in human beings, and the path forward exists. Phase II will test it in more people. Phase III will test whether it actually prevents infection. Those trials will take years.

But the idea that AI can design a vaccine from scratch, and that vaccine can work in a human body, is no longer theoretical. It happened this month. The world barely noticed. It should.