Various approaches have been proposed to forecast the maximum expected earthquake magnitude that may be induced by fluid injection in a given area. Proposed forecast methods include a geometrical approach based on inferred dimensions of the stimulated volume; a formula that predicts maximum magnitude based on a putative linear relationship between maximum seismic moment and net injected volume; and a probabilistic approach based on seismic-activity rate. In this study, the probabilistic approach is extended to include a tapered Gutenberg-Richter distribution, which accounts for the effects of finite-fault dimensions. Each method makes specific assumptions that impact the applicability of the maximum-magnitude forecast, leading to divergent implications for monitoring and mitigation. Starting from basic concepts from earthquake seismology, we outline the theory and applications of these forecasting methods and test the maximum-magnitude forecasts using published examples of induced earthquakes. The majority of published examples are consistent with the putative volumetric limit, but a number of anomalous hydraulic-fracturing-induced events suggest that maximum magnitude is ultimately limited by geology (i.e., fault dimensions) rather than operational factors (e.g., net injected volume). Progress in understanding maximum magnitude may contribute to improved public communication and a stronger scientific foundation for traffic light criteria.