I will argue that deep moral enhancement is relatively prone to unexpected consequences. I first argue that even an apparently straightforward example of moral enhancement such as increasing human co-operation could plausibly lead to unexpected harmful effects. Secondly, I generalise the example and argue that technological intervention on individual moral traits will often lead to paradoxical effects on the group level. Thirdly, I contend that insofar as deep moral enhancement targets higher-order desires, it is prone to be self-reinforcing and irreversible. (...) Fourthly, I argue that the complex causal history of moral traits, with its relatively high frequency of contingencies, indicates their fragility. Finally, I conclude that attempts at deep moral enhancement pose greater risks than other enhancement technologies. For example, one of the major problems that moral enhancement is hoped to address is lack of co-operation between groups. If humanity developed and distributed a drug that dramatically increased co-operation between individuals, we would likely see a paradoxical decrease in co-operation between groups and a self-reinforcing increase in the disposition to engage in further modifications – both of which are potential problems. (shrink)