Carbon‐14 radiotracer experiments at 10 kbar (1 GPa) and 1000°‐1200°C reveal that, due to low diffusivity or low solubility (or both), carbon is immobile in the grain boundary regions of natural dunite. Within the resolution of the beta‐track mapping technique, grain boundary diffusion of carbon from a pure graphite source is not detectable in experiments of up to 10 days duration at 1200°C, even though measurable volume diffusion profiles are generated at these conditions (D ∼ 3‐4 × 10−10cm²/s, in agreement with previous estimates). The documented immobility of carbon in dunite grain boundaries precludes development of a continuous carbon film by diffusion from an initially localized source. If such a film is the cause of the mantle high‐conductivity layer, it probably originated by a process involving fluid (CO/CO2) infiltration and su