Significance Statement Current theories suggest that accurately predicting the sensory consequences of one's actions is central for perception, awareness of action, and motor learning. In the latter case, it is assumed that prediction errors are used to train the controller that transforms our desired sensory consequences into motor commands. Here we show that, following exposure to biased hand visual feedback, people can update their ability to predict visual consequences of hand movements without necessarily improving their ability to control these movements. This work challenges the view that the joint update of prediction and control is mandatory when facing a change in the mapping between motor commands and sensory consequences. Instead, we propose that task demands mediate the update of prediction and control.