In cell and animal models, telomere erosion promotes chromosomal instability via breakage-fusion-bridge cycles, contributing to the early stages of tumorigenesis. However, evidence involving short telomeres in cancer development in humans is scarce, epidemiological and indirect. Here we directly implicate telomere shortening as a critical molecular event for malignant evolution in aplastic anemia (AA). Patients' telomere lengths at diagnosis of AA, while comparable to age-matched controls, inversely correlated with the probability of developing a cytogenetically abnormal clone. A significantly increased number of telomere signal-free chromosomal ends and chromosomal numerical and structural abnormalities were observed in bone marrow cells of patients with shorter telomeres in comparison with patients with longer telomeres and healthy subjects. The proportion of monosomy-7 cells in the bone marrow at diagnosis of AA inversely correlated with telomere length, years before the emergence of an autonomous and clinically detectable abnormal clone. Marrow cells of clinically healthy individuals carrying loss-of-function telomerase mutations and with extremely short telomeres also showed chromosomal instability in vitro. These results provide the first clinical direct evidence in humans that short telomeres in hematopoietic cells are dysfunctional, mediate chromosomal instability and predispose to malignant transformation in a human disease.