Gene silencing holds great promise for the treatment of several diseases and can be exploited to investigate gene function and activity of the regulatory genome. Here, we develop a novel modality of gene silencing that exploits epigenetics to achieve stable and highly efficient repression of target genes. To this end, we generated Artificial Transcriptional Repressors (ATRs), chimeric proteins containing a custom-made DNA binding domain fused to the effector domain of chromatin-modifying enzymes involved in silencing process of Endogenous RetroViruses (ERVs). By performing iterative rounds of selection in cells engineered to report for synergistic activity of candidate effector domains, we identified a combination of 3 domains (namely KRAB, DNMT3A and DNMT3L) that, when transiently co-assembled on the promoter of the reporter cassette, recreate a powerful embryonic-specific repressive complex capable of inducing full and long-term (>150 days) silencing of transgene expression in up to 90% of the cells. The ATR-induced silencing was cell type and locus independent, and resistant to metabolic activation of the cells. Importantly, these findings were holding true also for endogenous genes embedded in their natural chromatin context, as shown for the highly and ubiquitously expressed B2M gene. Here, transient co-delivery of TALE-based ATRs resulted in loss of surface expression of B2M and, consequently, of the MHC-I molecules in up to 80% of the cells. This phenotype was associated with a drastic switch in the epigenetic and transcriptional state of the constitutively active B2M promoter, which become highly decorated with de novo DNA methylation and deprived of RNAP II. Importantly, silencing was sharply confined to the targeted gene and resistant to INF-γ, a potent natural activator of B2M. We further extended these studies by showing that our silencing approach is portable to the CRISPR/dCas9 DNA binding technology. In this setting, comparable levels of B2M silencing (up to 80%) were achieved using either pools or even individual sgRNAs coupled to dCas9-based ATRs. Yet, adoption of this technology allowed performing simultaneous, highly efficient multiplex gene silencing within the same cell, as shown for B2M, IFNAR1 and VEGFA. Finally, we assessed resistance of the silenced gene to activity of potent artificial transcription activators and chromatin remodelers, and found that only targeted DNA demethylation was able to reawaken the silent gene. This allowed performing iterative cycles of silencing and reactivation of the same gene in the same cell population. Overall, these data provide the first demonstration of efficient and stable epigenetic silencing of endogenous genes upon transient delivery of ATRs. This was accomplished by repurposing the ERVs silencing machinery, which instructs self-sustaining repressive epigenetic states to the target gene. While silencing of B2M might be used to generate universally transplantable allogeneic cells, our hit-and-run strategy provides a powerful new alternative to conventional gene silencing for both basic and translational research.