AbstractWe present a multispecies stochastic model that suggests optimal fishing policy for two species in a three‐species predator–prey ecosystem in the Barents Sea. We employ stochastic dynamic programming to solve a three‐dimensional model, in which the catch is optimized by using a multispecies feedback strategy. Applying the model to the cod, capelin, and herring ecosystem in the Barents Sea shows that the optimal catch for the stochastic interaction model is more conservative than that implied by the deterministic model. We also find that stochasticity has a stronger effect on the optimal exploitation policy for prey (capelin) than for predator (cod).