The teratoma forming potential of pluripotent stem cells (PSCs), and genetic aberrations that may lead to tumor formation from PSC derivatives, are considered as a major safety risk for cellular therapies. Introduction of inducible suicide genes as synthetic fail-safe systems has been proposed to minimize these risks. Recent research challenged the usefulness of such systems even for targeted introduction via accurate gene editing approaches. Apparently transgene silencing and elimination of a HTK suicide gene through 'loss-of-heterozygosity' (LoH) led to cell clones that escaped the induced suicide. We have introduced an inducible Caspase9 (iCasp-9) suicide gene into induced pluripotent stem cells (iPSCs), that has already been applied clinically in other settings. The iCASP9 gene coupled to a red nuclear GFP variant under control of the CAG promoter was inserted into the AAVS1 locus, either monoallelic or homozygous on both alleles. Efficient induction of apoptosis in vitro could be induced via treatment of iCASP9 iPSCs with two chemical inducers of dimerization (CID) at different concentrations for 24 hours. While NODSCID mice after transplantation of undifferentiated monoallelic iCASP9 iPSCs under the kidney capsule developed teratomas, CID treatment for three days led to rapid shrinking of such tumor structures. In individual mice, however, that received monoallelic iCASP9 iPSCs, tumor-like human tissue could be detected after CID treatment. Further in vitro experiments confirmed that in very rare subclones monoallelic iCASP9 hiPSCs lost transgene expression and can became resistant to CID induction in vitro with frequencies of ~ 3x10^-8. Analysis of CID-resistant subclones identified either elimination of the transgene, presumably via LoH, or via methylation of the CAG promoter as underlying mechanism. In contrast, we never observed any CID resistant escapees form biallelic iCASP9 iPSC clones, even after treatment of up to 0,5x10^9 iPSCs. This observation further argues for LoH as underlying mechanism of transgene elimination in monoallelic clones and suggests that CAG promoter methylation on both alleles represent independent events. In conclusion, biallelic integration of an iCASP9 safety switch in the AAVS1 locus allows for efficient induction of cellular suicide and may substantially increase the safety level of iPSC-based therapies. We propose that safety levels should be calculated by relating the observed frequencies of clonal escapees to clinically relevant cell numbers, i.e. cell number in tumors of a size that is readily detectable by modern imaging approaches.