Oncogenic fusions involving transcription factors are present in the majority of pediatric leukemias; however, the context-specific mechanisms they employ to drive cancer remain poorly understood. CBFA2T3-GLIS2 (C/G) fusions occur in treatment-refractory acute myeloid leukemias and are restricted to young children. To understand how the C/G fusion drives oncogenesis we applied CUT&RUN chromatin profiling to an umbilical cord blood/endothelial cell (EC) co-culture model of C/G AML that recapitulates the biology of this malignancy. We find C/G fusion binding is mediated by its zinc finger domains. Integration of fusion binding sites in C/G- transduced cells with Polycomb Repressive Complex 2 (PRC2) sites in control cord blood cells identifies MYCN, ZFPM1, ZBTB16 and LMO2 as direct C/G targets. Transcriptomic analysis of a large pediatric AML cohort shows that these genes are upregulated in C/G patient samples. Single cell RNA-sequencing of umbilical cord blood identifies a population of megakaryocyte precursors that already express many of these genes despite lacking the fusion. By integrating CUT&RUN data with CRISPR dependency screens we identify BRG1/SMARCA4 as a vulnerability in C/G AML. BRG1 profiling in C/G patient-derived cell lines shows that the CBFA2T3 locus is a binding site, and treatment with clinically-available BRG1 inhibitors reduces fusion levels and downstream C/G targets including N-MYC, resulting in C/G leukemia cell death and extending survival in a murine xenograft model.