Imperfect knowledge of social–ecological systems can obscure predictability of resource fluctuation and, in turn, lead to erroneous risk assessments and delayed management actions. Systematic error in population status such as persistent overestimation of abundance is a pervasive conservation problem and has plagued assessments of commercial exploitation of marine species, threatening its sustainability.Using North Sea saithe (Pollachius virens)–a demersal (bottom-water) predatory fish–as a real-world case study, we illustrate a precautionary approach to diagnose robustness of harvest rules to persistent estimation bias (overestimated stock abundance and underestimated fishing mortality rate) and to develop alternative protective measures that minimize population depletion (quasi-extinction) risk by propagating known sources of uncertainty (process, observation, and implementation) through closed-loop simulation of resource–management feedback systems (management strategy evaluation, MSE).Analyses showed that the harvest rules set for saithe are robust to a moderate amount (10– 30%) of estimation bias. More severe bias sets overly optimistic catch limits and promotes overexploitation only in the short term; unacceptably high quasi-extinction risks, however, result primarily from progressively amplified amplitudes of catch fluctuation. Although these undesirable outcomes were, to some extent, mitigated by applying a policy tool to suppress catch fluctuation, this tool falls short of being an effective measure to achieve management goals.More consistent performance of management measures was achieved by developing and applying more precautionary harvest rules through MSE by explicitly accounting for bias. When bias became more severe, raising threshold abundance (by 8–24%) that triggers management actions and lowering target exploitation rate (by 6–29%) would not only safeguard against overexploitation and depletion but also provide catch stability (less disruption in fishing operations).We show that the precautionary approach to risk management through MSE offers a powerful tool to set safe harvest boundaries when assessments are persistently biased. Given challenges in identifying the sources, we suggest bias be routinely evaluated through MSE, and alternative measures be developed to set catch limits when needed. By explicitly accounting for key sources of uncertainty in managing commercial exploitation, our proposed approach ensures effective conservation and sustainable exploitation of living marine resources even under profound uncertainty.