We present the results from our non-LTE investigation for neutral carbon, which was carried out to remove potential systematic errors in stellar abundance analyses. The calculations were performed for late-type stars and give substantial negative non-LTE abundance corrections. When applied to observations of extremely metal-poor stars, which within the LTE framework seem to suggest a possible [C/O] uprise at low metallicities (Akerman et al. 2004), these improvements will have important implications, enabling us to understand if the standard chemical evolution model is adequate, with no need to invoke signatures by Pop. III stars for the carbon nucleosynthesis.