Preventive therapy remains key to the elimination of tuberculosis and is typically offered to people with presumptive Mycobacterium tuberculosis infection to prevent active disease. Although the duration of tuberculosis preventive therapy has been reduced substantially over time, it remains long in absolute terms, and uptake remains low. Treatment-shortening trials using non-inferiority designs have so far led to the implementation of effective regimens of 1–4 months’ duration. Such regimens are a substantial improvement on the previous 6–9 months’ duration standard of care but still far too long given potential toxicity and the very low baseline risk of disease for most individuals. The efficacy of even shorter tuberculosis preventive therapy regimens, including ultra-short regimens shorter than 2 weeks’ duration, is yet to be explored, but optimal public health outcomes might be achieved even if the efficacy of such regimens is lower than that of the standard of care. Greater acceptability could lead to higher population uptake, and, potentially, to more cases of tuberculosis avoided. Nonetheless, the optimal duration of ultra-short tuberculosis preventive therapy regimens cannot be explored through classic two-arm non-inferiority trials. Instead, the relationship between different durations and efficacy of tuberculosis preventive therapy will need to be characterised, requiring some participants to be randomly assigned to no (or delayed) therapy in order to characterise the number of tuberculosis cases averted by the shortest options. We argue that such trials are needed to identify the optimal trade-off between efficacy and acceptability and would be ethically acceptable provided there were appropriate risk mitigation measures for participants, including careful monitoring for the development of active disease. In this Personal View, we discuss some of the scientific and ethical considerations around the investigation of ultra-short-course preventive therapy for tuberculosis.