Spatiotemporal control over the release or activation of biomacromolecules such as siRNA remains a significant challenge. Light-controlled release has gained popularity in recent years; however, a major limitation is that most photoactivable compounds/systems respond only to UV irradiation, but not near-infrared (NIR) light that offers a deeper tissue penetration depth and better biocompatibility. This paper reports a simple NIR-to-UV upconversion nanoparticle (UCNP)-based siRNA nanocarrier for NIR-controlled gene silencing. siRNA is complexed onto a NaYF4:Yb/Tm/Er UCNP through an azobenzene (Azo)–cyclodextrin (CD) host–guest interaction. The UV emission generated by the NIR-activated UCNP effectively triggers the trans-to-cis photoisomerization of azobenzene, thus leading to the release of siRNA due to unmatched host – guest pairs. The UCNP-siRNA complexes are also functionalized with PEG (i.e., UCNP-(CD/Azo)-siRNA/PEG NPs), targeting ligands (i.e., EGFR-specific GE11 peptide), acid-activatable cell-penetrating peptides (i.e., TH peptide), and imaging probes (i.e., Cy5 fluorophore). The UCNP-(CD/Azo)-siRNA/PEG NPs with both GE11 and TH peptides display a high level of cellular uptake and an excellent endosomal/lysosomal escape capability. More importantly, NIR-controlled spatiotemporal knockdown of GFP expression is successfully achieved in both a 2D monolayer cell model and a 3D multicellular tumor spheroid model. Thus, this simple and versatile nanoplatform has great potential for the selective activation or release of various biomacromolecules.