An artificial intelligence-designed universal coronavirus vaccine has passed its first human clinical trial with no serious safety signals and robust immune responses across all 72 participants, according to results published Monday in The Lancet. The data marks the first time a computationally engineered vaccine candidate targeting the broad betacoronavirus family has cleared a Phase 1 threshold in humans — a result that public health researchers said opens a credible path toward pandemic preparedness tools that don't require waiting for a novel pathogen to emerge before vaccine design can begin.
How the AI Built It
The vaccine was designed using a generative protein modeling platform developed at the National Institutes of Health's National Institute of Allergy and Infectious Diseases in Bethesda, Maryland, in collaboration with researchers at the University of Washington's Institute for Protein Design. Rather than targeting the spike protein of a single known strain — the approach used in the COVID-19 vaccines deployed beginning in 2021 — the AI system was trained to identify conserved epitopes shared across dozens of betacoronavirus variants, including SARS-CoV-2, MERS-CoV, and several bat-origin lineages considered likely precursors of future pandemic strains.
The resulting immunogen is a mosaic nanoparticle: a spherical protein scaffold that simultaneously presents fragments from eight different coronavirus lineages. The computational design process took approximately six weeks — compared to the one to two years typically required for conventional structure-based vaccine engineering. The speed advantage comes not only from raw computational throughput but from the AI's ability to explore a vastly larger design space than human researchers working through manual structure prediction.
"The AI didn't just compress the timeline," said a principal investigator on the study, speaking on condition of anonymity because the research team is under media embargo ahead of full journal publication. "It found epitope combinations that human designers would not have prioritized. The cross-variant coverage is genuinely broader than anything we've produced by hand."
Trial Results
The Phase 1 trial was conducted across three clinical research sites, including one in Rockville, Maryland, and enrolled 72 healthy adult volunteers who received two doses of the vaccine 28 days apart. All 72 participants generated neutralizing antibodies against both SARS-CoV-2 and MERS-CoV, according to the published data. Roughly 65% generated measurable neutralization responses against at least four additional betacoronavirus strains they had never been naturally exposed to — a key benchmark for the "universal" designation researchers are cautiously applying to the candidate.
No participants developed serious adverse events. Mild injection-site reactions and transient fatigue — standard responses to mRNA-based platforms — were reported in approximately 40% of participants, consistent with the profiles seen in licensed influenza and COVID-19 vaccines. Researchers noted comparable immune response profiles across age groups within the 18-to-55 range enrolled in Phase 1, though elderly participants were excluded under standard age stratification protocols and will be included in Phase 2.
Phase 2 trials, designed to test efficacy in larger international populations, are expected to begin enrollment in Q4 2026, according to ClinicalTrials.gov registration data. Regulatory submissions to the Food and Drug Administration and European Medicines Agency are not expected before 2028 at the earliest, even under an accelerated development schedule.
Why It Matters for Pandemic Preparedness
Public health researchers have argued for more than a decade that the reactive model of vaccine development — engineering a new product after a novel pathogen emerges and manufacturing it under crisis conditions — leaves too narrow a window to prevent significant early mortality in a fast-moving outbreak. COVID-19 killed an estimated 7 million people globally before vaccines were widely available, according to the World Health Organization, with the overwhelming share of deaths occurring in the 12 months before widespread immunization was achieved.
A vaccine that provides cross-protective immunity against a broad family of related viruses could, in theory, be pre-positioned in global stockpiles and administered at the first sign of a novel coronavirus outbreak — before the new strain's full genetic sequence is characterized in enough detail to design a strain-specific product. Whether the current candidate delivers on that promise depends entirely on Phase 2 efficacy results, but immunologists contacted by The Associated Press described the Phase 1 data as "genuinely encouraging, though not yet definitive."
Beyond Coronaviruses
The broader significance of Monday's results may lie less in the specific vaccine candidate than in what it validates about AI-driven protein design as a credible clinical development tool. Several research groups at the NIH, MIT, and the Scripps Research Institute are applying similar computational approaches to influenza, respiratory syncytial virus, and HIV — pathogens that have historically resisted conventional vaccine strategies because of their high mutation rates or sophisticated immune evasion mechanisms.
The Gates Foundation, which co-funded the research alongside the NIH and the Wellcome Trust, said in a statement that it views the trial results as a proof of concept for a platform-first approach to pandemic preparedness — developing and validating the design tools in advance rather than building them from scratch under crisis conditions. "The question is no longer whether AI can design effective vaccines," the statement read. "The question is how quickly we can build the regulatory and manufacturing infrastructure around this capability." Answering that question, as always in drug development, is the harder and considerably longer journey.