Background. New, simple, and better-performing diagnostic tools are needed for the diagnosis of tuberculosis (TB). Much effort has been invested in developing an antibody-based test for TB, but to date, no such test has performed with sufficient sensitivity and specificity. A key question remaining is the extent to which the disappointing performance of current tests is associated with a high background prevalence of latent TB. Methods. We compared Mycobacterium tuberculosis-specific ESAT-6 and CFP-10 antibody responses in a total of 565 human serum samples from M. tuberculosis-uninfected donors and donors with latent infection, as well as samples from patients with active TB. Our study included samples from 4 countries, representing environments with low, intermediate, and high TB incidences. Results. We demonstrated significant increases in antibody levels in latently infected contacts, compared with M. tuberculosis-uninfected individuals, and in patients with active TB disease, compared with latently infected contacts. Furthermore, we found a striking increase in the magnitude of the antibody responses in samples obtained from infected Ethiopian individuals (with and without disease), compared with Danish and Brazilian infected individuals; this was presumably the result of higher exposure levels. Conclusions. Our study confirms the presence of ESAT-6 and CFP-10 antibodies in patients with TB, and we demonstrate that significant antibody responses are not restricted to active TB disease but can reflect latent infection, particularly in areas with high levels of exposure to M. tuberculosis. This finding is important for the understanding of the poor discriminatory power of current serodiagnostic tests in regions of endemicity, and it may have major implications on the future development of serologic tests.