Given the high criminal recidivism rate, there is a pressing need to find new approaches to offender rehabilitation (Ministry of Justice, 2013). Recently, there has been increasing academic discussion of using biomedical interventions in rehabilitation programmes (see, e.g. Birks and Douglas, 2018). The discussion of biomedical criminal rehabilitation intersects with the literature on “moral bioenhancement” – the idea that biomedical interventions could be used in order to morally improve individuals' motivations, decision-making and conduct (and to decrease the likelihood of seriously immoral actions, including criminal behaviour). There is a growing consensus among advocates of moral bioenhancement that the most defensible type of moral enhancement would involve enhancing cognitive capacities and enhancing control over one's thought-processes, emotions and behaviour. This article will discuss a potential problem with attempting to use cognitive enhancement to rehabilitate offenders - a problem that, so far, has not been the subject of detailed analysis. Cognitive enhancement is a double-edged sword. The very same cognitive capacities that are required for moral reasoning, and which would be targeted by the envisaged enhancement technologies, are also the capacities that could be used for criminal behaviour. Therefore, there is a risk that cognitive enhancement might be counterproductive and increase the likelihood of reoffending. Enhancing the cognitive skills of badly motived people may have a greater chance of making them more dangerous than of making them more moral. If the double-edged sword problem cannot be overcome, it would seriously undermine the case for moral bioenhancement. It would be a severe restriction on the usefulness of moral bioenhancement if the most promising means of achieving it could only be given to those who are least in need of moral enhancement – those who are already well motivated.