Researchers from the University of Tokyo developed an AI tool (AI‑IR) that predicts insulin resistance from nine routine clinical measures and applied it to 500,000 UK Biobank participants, finding insulin resistance is a risk factor for 12 types of cancer. Their results, published in Nature Communications, provide population‑scale evidence linking metabolic dysfunction to cancer incidence. The study demonstrates how machine learning can approximate hard‑to‑measure metabolic states at scale and suggests insulin resistance may be a modifiable cancer risk marker; it also raises potential for integrating predicted metabolic phenotypes into cancer risk stratification.