Containment of Mycobacterium tuberculosis (Mtb) infection requires T cell recognition of infected macrophages. Mtb has evolved to tolerate, evade, and subvert host immunity. Despite a vigorous and sustained CD8+ T cell response during Mtb infection, CD8+ T cells make limited contribution to protection. Here, we ask whether the ability of Mtb-specific T cells to restrict Mtb growth is related to their capacity to recognize Mtb-infected macrophages. We derived CD8+ T cell lines that recognized the Mtb immunodominant epitope TB10.44−11 and compared them to CD4+ T cell lines that recognized Ag85b240-254 or ESAT63-17. While the CD4+ T cells recognized Mtb-infected macrophages and inhibited Mtb growth in vitro, the TB10.4-specific CD8+ T cells neither recognized Mtb-infected macrophages nor restricted Mtb growth. TB10.4-specific CD8+ T cells recognized macrophages infected with Listeria monocytogenes expressing TB10.4. However, over-expression of TB10.4 in Mtb did not confer recognition by TB10.4-specific CD8+ T cells. CD8+ T cells recognized macrophages pulsed with irradiated Mtb, indicating that macrophages can efficiently cross-present the TB10.4 protein and raising the possibility that viable bacilli might suppress cross-presentation. Importantly, polyclonal CD8+ T cells specific for Mtb antigens other than TB10.4 recognized Mtb-infected macrophages in a MHC-restricted manner. As TB10.4 elicits a dominant CD8+ T cell response that poorly recognizes Mtb-infected macrophages, we propose that TB10.4 acts as a decoy antigen. Moreover, it appears that this response overshadows subdominant CD8+ T cell response that can recognize Mtb-infected macrophages. The ability of Mtb to subvert the CD8+ T cell response may explain why CD8+ T cells make a disproportionately small contribution to host defense compared to CD4+ T cells. The selection of Mtb antigens for vaccines has focused on antigens that generate immunodominant responses. We propose that establishing whether vaccine-elicited, Mtb-specific T cells recognize Mtb-infected macrophages could be a useful criterion for preclinical vaccine development. Immunodominant antigens elicit a large majority of T cells during an infection, and it is presumed that these T cells go on to recognize infected cells. Immunodominant antigens produced by Mycobacterium tuberculosis (Mtb) have been incorporated into vaccines, but whether T cells specific for these antigens recognize Mtb-infected cells is inconsistent. One of these is TB10.4 (EsxH), and after aerosol infection in mice, up to 40% of lung CD8 T cells recognize TB10.4. Vaccination with TB10.44−11 peptide elicits a robust response and TB10.4-specific memory CD8 T cells develop. However, these mice are not protected against Mtb challenge. In trying to understand why this vaccine fails, we discovered that TB10.4-specific CD8 T cells do not recognize Mtb-infected macrophages in vitro. In contrast, under identical conditions, Ag85b-specific CD4 T cells recognize Mtb-infected macrophages and inhibit Mtb growth. In contrast, polyclonal CD4 and CD8 T cells from the lungs of infected mice recognize Mtb-infected macrophages, suggesting macrophages present antigens other than the immunodominant TB10.4 antigen. Thus, we conclude that TB10.4 is a decoy antigen that elicits a dominant CD8 T cell response that poorly recognizes Mtb-infected macrophages, and allows Mtb to evade CD8 immunity. Instead, this response may benefit Mtb by promoting inflammation. We propose that the poor ability of CD8 T cells to mediate protection might arise because the dominant T cells are unable to recognize infected macrophages. Identifying antigens that are presented by infected cells should be prioritized, as their inclusion in vaccines may enhance the expansion of T cells that mediate protection.