Inhibition of prepotent action is an important aspect of self-control, particularly in social contexts. Action inhibition and its neural bases have been extensively studied. However, the neural precursors of free decisions to inhibit have hardly been studied. We asked participants to freely choose to either make a rapid key press in response to a visual cue, or to transiently inhibit action, and briefly delay responding. The task required a behavioural response on each trial, so trials involving inhibition could be distinguished from those without inhibition as those showing slower reaction times. We used this criterion to classify free-choice trials as either rapid or inhibited/delayed. For 13 participants, we measured the mean amplitude of the ERP activity at electrode Cz in three subsequent 50 ms time windows prior to the onset of the signal that either instructed to respond or inhibit, or gave participants a free choice. In two of these 50 ms time windows (-150 to -100, and -100 to -50 ms relative to action onset), the amplitude of prestimulus ERP differed between trials where participants "freely" chose whether to inhibit or to respond rapidly. Larger prestimulus ERP amplitudes were associated with trials in which participants decided to act rapidly as compared to trials in which they decided to delay their responses. Last-moment decisions to inhibit or delay may depend on unconscious preparatory neural activity.