The recent classification of curcumin (Cur) as a pan-assay interference compound (PAINS) and an invalid metabolic panaceas (IMPS) candidate demonstrated the controversial nature of Cur as a drug lead owing to its aggregation in aqueous phase and inherent instability in vivo. Here, we report a simple prodrug approach to generate nanoparticles of Curin situ that allow it to function reproducibly as an anticancer and an anti-inflammatory agent. Diphosphorylated curcumin (Cur-2p), a precursor of Cur and a substrate of alkaline phosphatase (ALP), exhibited drastically improved chemical stability and low aggregation in water. After conversion to curcumin around or inside cancer cells by ALP, Cur-2p selectively inhibited cancer cells that overexpressed ALP, but did not affect normal cells. Moreover, the intravitreal injection of Cur-2p resulted in excellent intraocular biocompatibility with no apparent damage to the morphology and visual function of retina, as shown by fundus imaging, optical coherence tomography (OCT), and histological observation. A rodent model of uveitis showed that Cur-2p significantly suppressed the inflammation response compared with Cur. As a rational approach to investigate and apply PAINS and IMPS candidates, this work presents a straightforward method to maximize the potential of drug leads and ultimately fulfil the promises and potential clinical benefits of PAINS and IMPS candidates.