AI Predicts Diabetes 12 Years Early in Major Breakthrough

Imagine knowing your diabetes risk 12 years before symptoms appear. That future just became possible thanks to a groundbreaking AI system from Israeli researchers.

The AI model, called GluFormer, can forecast type 2 diabetes up to 12 years in advance by analyzing patterns in continuous glucose monitoring data. Published this week in Nature, the research was led by teams at Nvidia Israel, the Weizmann Institute of Science, Pheno.AI, and Mohamed bin Zayed University of Artificial Intelligence.

What makes this truly remarkable is how little data the system needs. Sometimes just one to two weeks of glucose readings can predict health outcomes more than a decade away.

The model was trained using Israel's "10K Project," one of the world's most comprehensive health datasets. More than 14,000 participants were tracked for years, combining glucose measurements with genetic testing, microbiome profiling, sleep studies, and detailed medical histories.

Professor Eran Segal, a computational biologist at the Weizmann Institute, explained the problem this tool solves. "People classified as pre-diabetic are often treated as a single group, but existing measures do not reliably predict who will actually go on to develop diabetes," he said. "The AI model can."

The results speak for themselves. GluFormer flagged 66 percent of participants who later developed diabetes as high-risk years before diagnosis. Only a tiny fraction of those marked low-risk went on to develop the disease.

The system's abilities extend beyond diabetes. It identified nearly 70 percent of individuals who later died from heart-related causes as high-risk. The model also predicted outcomes linked to kidney and liver function, blood lipid levels, visceral fat, and sleep disorders.

Guy Lutsker, an AI researcher at Nvidia, said the system appears to have learned fundamental aspects of metabolic disease that simple clinical rules miss. "Like many powerful AI systems, it functions partly as a black box, but it delivers predictions that go well beyond what existing methods can achieve," he explained.

The Ripple Effect

The implications reach far beyond individual patients. Diabetes prevention is a major priority for health systems worldwide, particularly in the United States where pre-diabetes interventions cost billions annually. The global cost of diabetes is projected to hit $2.5 trillion by 2030.

By pinpointing which patients face the highest risk, healthcare providers can direct resources where they matter most. Lifestyle programs, monitoring, and drug treatments can go to people who will truly benefit rather than being spread across everyone with slightly elevated blood sugar.

Pheno.AI holds the rights to commercialize the technology and plans to work with healthcare organizations to bring GluFormer into clinical use. The study highlights Israel's growing role at the intersection of artificial intelligence, healthcare, and big data.

For Segal, the mission is clear. "Only a minority of people defined as pre-diabetic will actually develop diabetes. The challenge is knowing who they are, and now we're much closer to answering that."