Traditional safety performance requirements for nuclear reactors have been developed for critical reactors, whose kinetics characteristics differ significantly from sub-critical, accelerator-driven nuclear reactors. In a critical nuclear reactor, relatively small amounts of reactivity (negative or positive) can produce large changes in the fission rate. In sub-critical reactors, the self-multiplication (k) decreases as the sub-criticality (1-k) increases, and the responsiveness to small reactivity changes decreases. This makes sub-critical nuclear reactors less responsive to positive reactivity insertions than critical reactors. Also, larger negative reactivity insertions are needed in sub-critical reactors to shut down the fission chain if the neutron source remains. This paper presents the results from a computational analysis of the safety performance of sub-critical, accelerator-driven nuclear reactors. Coupled kinetics and thermal-hydraulics models are used to quantify the effectiveness of traditional protection and control system designs in sub-critical reactors. The analyses also quantify the role of inherent, passive reactivity feedback mechanisms in sub-critical reactors. Computational results are used to develop conclusions regarding the most favorable and effective means for reactor control and protection in sub-critical, accelerator-driven nuclear reactors.