Abstract Future projections of the European climate suffer from uncertainties in the changes of the North Atlantic jet stream. Most climate models simulate an annual-mean poleward shift of the jet in response to global warming, but the responses have a large spread. Previous studies comparing climate models suggest that the projected meridional shift of the jet is negatively correlated with its simulated present-day mean latitude. Model basic-state biases are a possible cause for the uncertainty, but their effect is difficult to assess because the spread in simulations may be caused by any intermodel differences. In this study, the effect of model biases on future projections was isolated by modifying the basic state of a single atmospheric model with a run-time correction method aiming to adjust the model climatology toward those of three other models and a reanalysis. The effect of model biases was found to be strongly seasonal. In winter, changes in the frequencies of two of the three preferred positions of the jet were found to be sensitive to the model biases, causing considerable uncertainty in the jet shift and contributing to the anticorrelation between the present-day jet latitude and the future shift, but in summer, the impact of biases is small relative to the magnitude of the changes. While the anticorrelation of jet latitude and shift is only partly caused by biases, our results demonstrate there is potential to reduce uncertainty in the projected jet stream changes by improving model basic states. Significance Statement Future projections of climate models are affected by biases, i.e., systematic errors, in simulations of the present-day climate. We investigated the contribution of model biases to the large spread of simulated future changes of the North Atlantic jet stream, a varying current of fast winds strongly affecting the European climate. We found that biases cause large uncertainty in, especially, the changes of the position of the winter jet stream. The results also show that a known dependence of the latitudinal shift of the jet on its simulated current position is not as strong as suggested by previous studies. However, reducing model biases provides opportunities for more reliable climate predictions.