The global death toll from tuberculosis (TB) is an ongoing tragedy (1). Currently, clinicians in TB-prevalent resource-poor settings attempt to control the pandemic with clearly inadequate tools: bacillus Calmette-Guérin (BCG) vaccination, sputum smear microscopy, and antibiotic regimens of a minimum of 6 months, each of which has remained essentially unchanged for many decades (2). Although initial trials of novel vaccinations have been disappointing (3, 4), two recent vaccine trials have generated more promising results (5, 6). The M72/AS01E vaccine reduced the number of active cases by 50% in a phase 2b study (5), whereas a repeat BCG administration reduced sustained IFN-g release assay conversion from 11.6% to 6.7%(6). Repeat BCG was not included in the first study, and so the relative efficacy cannot be defined. However, both studies demonstrate that a significant residual disease burden will persist, which can lead to ongoing transmission. Consequently, it is vital to consider what constitutes a protective immune response to Mycobacterium tuberculosis (Mtb) versus a pathological immune response that leads to cavitation and transmission (7). We review clinical and experimental observations that highlight the complexity of the host–pathogen interaction in human TB to develop an entirely new conceptual model that will inform future strategies.