Kimberlites of the Udachnaya-East pipe (Siberia) include a uniquely dry and serpentine-free rock type with anomalously high contents of chlorine (Cl ≤ 6.1 wt%), alkalies (Na2O + K2O ≤ 10 wt%) and sulfur (S ≤ 0.50 wt%), referred to as a “salty” kimberlite. The straightforward interpretation is that the Na-, K-, Cl- and S-rich components originate directly from a carbonate-chloride kimberlitic magma that is anhydrous and alkali-rich. However, because brines and evaporites are present on the Siberian craton, previous studies proposed that the kimberlitic magma was contaminated by the assimilation of salt-rich crustal rocks. To clarify the origin of high Cl, alkalies and S in this unusual kimberlite, here we determine its sulfur speciation and isotopic composition and compare it to that of non-salty kimberlites and kimberlitic breccia from the same pipe, as well as potential contamination sources (hydrothermal sulfides and sulfates, country-rock sediment and brine collected in the area). The average δ34S of sulfides is − 1.4 ± 2.2‰ in the salty kimberlite, 2.1 ± 2.7‰ in the non-salty kimberlites and 14.2 ± 5.8‰ in the breccia. The average δ34S of sulfates in the salty kimberlites is 11.1 ± 1.8‰ and 27.3 ± 1.6‰ in the breccia. In contrast, the δ34S of potential contaminants range from 20 to 42‰ for hydrothermal sulfides, from 16 to 34‰ for hydrothermal sulfates, 34‰ for a country-rock sediment (Chukuck suite) and the regional brine aquifer. Our isotope analyses show that (1) in the salty kimberlites, neither sulfates nor sulfides can be simply explained by brine infiltration, hydrothermal alteration or the assimilation of known salt-rich country rocks and instead, we propose that they are late magmatic phases; (2) in the non-salty kimberlite and breccia, brine infiltration lead to sulfate reduction and the formation of secondary sulfides – this explains the removal of salts, alkali-carbonates and sulfates, as well as the minor olivine serpentinization; (3) hydrothermal sulfur was added to the kimberlitic breccia, but not to the massive kimberlites. In situ measurements of sulfides confirm this scenario, clearly showing the addition of two sulfide populations in the breccia (pyrite-pyrrhotites with average δ34S of 7.9 ± 3.4‰ and chalcopyrites with average δ34S of 38.0 ± 0.4‰) whereas the salty and non-salty kimberlites preserve a unique population of djerfisherites (Cl- and K-rich sulfides) with δ34S values within the mantle range. This study provides the first direct evidence of alkaline igneous rocks in which magmatic sulfate is more abundant than sulfide. Although sulfates have been rarely reported in mantle materials, sulfate-rich melts may be more common in the mantle than previously thought and could balance the sulfur isotope budget of Earth's mantle.