Cellular senescence induced by chemotherapy leads to the acquisition of stemness in cancer cells, which results in enhanced tumour-promoting capacity after forced release or spontaneous escape from the senescent cell-cycle arrest. Cells can go into senescence under stressed conditions to prevent the spread of potentially cancerous cells. Clemens Schmitt and colleagues show that cellular senescence induced by chemotherapy leads to the acquisition of 'stemness' in cancer cells, which allows them to escape senescence and promote tumour growth. Interestingly, such cells gain elevated tumour-initiating capacity compared with cells that have never undergone senescence. Cellular senescence is a stress-responsive cell-cycle arrest program that terminates the further expansion of (pre-)malignant cells1,2. Key signalling components of the senescence machinery, such as p16INK4a, p21CIP1 and p53, as well as trimethylation of lysine 9 at histone H3 (H3K9me3), also operate as critical regulators of stem-cell functions (which are collectively termed ‘stemness’)3. In cancer cells, a gain of stemness may have profound implications for tumour aggressiveness and clinical outcome. Here we investigated whether chemotherapy-induced senescence could change stem-cell-related properties of malignant cells. Gene expression and functional analyses comparing senescent and non-senescent B-cell lymphomas from Eμ-Myc transgenic mice revealed substantial upregulation of an adult tissue stem-cell signature, activated Wnt signalling, and distinct stem-cell markers in senescence. Using genetically switchable models of senescence targeting H3K9me3 or p53 to mimic spontaneous escape from the arrested condition, we found that cells released from senescence re-entered the cell cycle with strongly enhanced and Wnt-dependent clonogenic growth potential compared to virtually identical populations that had been equally exposed to chemotherapy but had never been senescent. In vivo, these previously senescent cells presented with a much higher tumour initiation potential. Notably, the temporary enforcement of senescence in p53-regulatable models of acute lymphoblastic leukaemia and acute myeloid leukaemia was found to reprogram non-stem bulk leukaemia cells into self-renewing, leukaemia-initiating stem cells. Our data, which are further supported by consistent results in human cancer cell lines and primary samples of human haematological malignancies, reveal that senescence-associated stemness is an unexpected, cell-autonomous feature that exerts its detrimental, highly aggressive growth potential upon escape from cell-cycle blockade, and is enriched in relapse tumours. These findings have profound implications for cancer therapy, and provide new mechanistic insights into the plasticity of cancer cells.