Generation of reactive oxygen species (ROS) is the hallmark of important biological processes and photodynamic therapy (PDT), where ROS production results from in situ illumination of certain dyes. Here we test the hypothesis that the yield, fate, and efficacy of the species evolved highly depend on the dye's environment. We show that Pd-bacteriopheophorbide (Pd-Bpheid), a useful reagent for vascular targeted PDT (VTP) of solid tumors, which has recently entered into phase II clinical trials under the code name WST09 (trade name TOOKAD), forms appreciable amounts of hydroxyl radicals, superoxide radicals, and probably hydrogen peroxide in aqueous medium but not in organic solvents where singlet oxygen almost exclusively forms. Evidence is provided by pico- and nanosecond time-resolved spectroscopies, ESR spectroscopy with spin-traps, time-resolved singlet oxygen phosphorescence, and chemical product analysis. The quantum yield for singlet oxygen formation falls from similar to 1 in organic solvents to similar to 0.5 in membrane-like systems (micelles or liposomes), where superoxide and hydroxyl radicals form at a minimal quantum yield of 0.1%. Analysis of photochemical products suggests that the formation of oxygen radicals involves both electron and proton transfer from Pd-3-Bpheid at the membrane/water interface to a colliding oxygen molecule, consequently forming superoxide, then hydrogen peroxide, and finally hydroxyl radicals, with no need for metal catalysis. The ability of bacteriochlorophyll (Bchl) derivatives to form such radicals upon excitation at the near infrared (NIR) domain opens new avenues in PDT and research of redox regulation in animals and plants.