A number of writers have recently expressed concern that brain-computer interfaces may fail to interpret the intentions of the user accurately. In those cases, the user cannot be said to have been in control of their actions, and are therefore not culpable for them. This is taken to imply that a person is only culpable for a BCI-mediated action if the BCI interpreted the neural activity appropriately. In this talk, I argue this rests on a confusion about how BCIs work. There is no need for a BCI to interpret a person’s intentions accurately because BCIs do not work by means of interpreting intentions. I also argue for a more positive conclusion according to which a BCI-mediated action should be evaluated according to standards with which we evaluate actions carried out by means of ordinary tools. Whenever an action is performed by means of a complex tool, the output of which depends on many degrees of freedom inherent in its design, moral assessment of the action is difficult, and uncertain. BCI-mediated action is like this. Moral uncertainty is due to the complexity of the tool, rather than the fact that it interfaces with the brain.