Next-generation Trop-2–targeted therapy against advanced cancers is hampered by expression of Trop-2 in normal tissues. We discovered that Trop-2 undergoes proteolytic activation by ADAM10 in cancer cells, leading to the exposure of a previously inaccessible protein groove flanked by two N-glycosylation sites. We designed a recognition strategy for this region, to drive selective cancer vulnerability in patients. Most undiscriminating anti–Trop-2 mAbs recognize a single immunodominant epitope. Hence, we removed it by deletion mutagenesis. Cancer-specific, glycosylation-prone mAbs were selected by ELISA, bio-layer interferometry, flow cytometry, confocal microscopy for differential binding to cleaved/activated, wild-type and glycosylation site–mutagenized Trop-2. The resulting 2G10 mAb family binds Trop-2–expressing cancer cells, but not Trop-2 on normal cells. We humanized 2G10 by state-of-the-art complementarity determining region grafting/re-modeling, yielding Hu2G10. This antibody binds cancer-specific, cleaved/activated Trop-2 with Kd < 10−12 mol/L, and uncleaved/wtTrop-2 in normal cells with Kd 3.16×10−8 mol/L, thus promising an unprecedented therapeutic index in patients. In vivo, Hu2G10 ablates growth of Trop-2–expressing breast, colon, prostate cancers, but shows no evidence of systemic toxicity, paving the way for a paradigm shift in Trop-2–targeted therapy.