Using exchange Monte Carlo (MC) simulations based on an ab initio-parameterized Cluster Expansion (CE) model, we explore the phase stability of low-Cr Fe-Cr alloys as a function of vacancy (Vac), carbon, and nitrogen content. To parameterize the CE model, we perform Density Functional Theory calculations for more than 1600 supercells containing Cr-Vac-C-N clusters of various sizes in pure bcc Fe, Cr, and Fe-Cr alloys. MC simulations performed for T=650 K show that Cr clustering in Fe-3.28 at.%Cr alloys does not occur if there are no defects or if only vacancies are present. But the addition of a small amount of C or N, at the level as low as 0.02 at.% in an alloy with no vacancies, routinely results in the formation of ordered compounds containing a high amount of Cr, C and N. Cr segregates to interstitial atoms and Cr content in such Cr-rich clusters increases as a function of C and/or N concentration. In the presence of vacancies, C/N aggregate to the core regions of vacancy clusters, making segregation of Cr-rich clusters less pronounced. The structure of Cr-rich clusters varies significantly, depending on the concentration of interstitial atoms and on the ratio of N to C. Predictions derived from MC simulations agree with experimental observations of Fe-Cr alloys exposed to ion irradiation. The concentration of Cr found in clusters containing C and N interstitial atoms is in qualitative agreement, and the absolute Cr content found in the clusters simulated at 650 K is in quantitative agreement with experimental Atom Probe Tomography (APT) observations of Fe-3.28 at.%Cr alloys irradiated at 623 K. The measured C and N content of 42±5 and 151±3 atomic ppm likely results from the contamination that occurred during ion beam irradiation.