RUNX1 is generally considered a tumor suppressor in myeloid neoplasms. Inactivating RUNX1 mutations have frequently been found in patients with myelodysplastic syndrome (MDS) and cytogenetically normal acute myeloid leukemia (AML). However, no somatic RUNX1 alteration was found in AMLs with leukemogenic fusion proteins, such as core-binding factor (CBF) leukemia and MLL fusion leukemia, raising the possibility that RUNX1 could actually promote the growth of these leukemia cells. Using normal human cord blood cells and those expressing leukemogenic fusion proteins, we discovered a dual role of RUNX1 in myeloid leukemogenesis. RUNX1 overexpression inhibited the growth of normal cord blood cells by inducing myeloid differentiation, whereas a certain level of RUNX1 activity was required for the growth of AML1-ETO and MLL-AF9 cells. Using a mouse genetic model, we also showed that the combined loss of Runx1/Cbfb inhibited leukemia development induced by MLL-AF9. RUNX2 could compensate for the loss of RUNX1. The survival effect of RUNX1 was mediated by BCL2 in MLL fusion leukemia. Our study unveiled an unexpected prosurvival role for RUNX1 in myeloid leukemogenesis. Inhibiting RUNX1 activity rather than enhancing it could be a promising therapeutic strategy for AMLs with leukemogenic fusion proteins.