Recently, metal particle polymer composites have been proposed as sensing materials for micro corrosion sensors. To design the sensors, a detailed understanding of diffusion through metal particle polymer composites is necessary. Accordingly, in this work molecular dynamics (MD) simulations are used to study the diffusion of O2 and N2 penetrants in metal particle polymer nanocomposites composed of an uncross-linked polydimethylsiloxane (PDMS) matrix with Cu nanoparticle inclusions. PDMS is modeled using a hybrid interatomic potential with explicit treatment of Si and O atoms along the chain backbone and coarse-grained methyl side groups. In most models examined in this work, MD simulations show that diffusion coefficients of O2 and N2 molecules in PDMS-based nanocomposites are lower than that in pure PDMS. Nanoparticle inclusions act primarily as geometric obstacles for the diffusion of atmospheric penetrants, reducing the available porosity necessary for diffusion, with instances of O2 and N2 molecule trapping also observed at or near the PDMS/Cu nanoparticle interfaces. In models with the smallest gap between Cu nanoparticles, MD simulations show that O2 and N2 diffusion coefficients are higher than that in pure PDMS at the lowest temperatures studied. This is due to PDMS chain confinement at low temperatures in the presence of the Cu nanoparticles, which induces low-density regions within the PDMS matrix. MD simulations show that the role of temperature on diffusion can be modeled using the Williams–Landel–Ferry equation, with parameters influenced by nanoparticle content and spacing.