The idea that people’s numerical identity may sometimes be discontinuous over time initially appears to provide useful material for defending restrictions on putatively self-harming behaviour in a non-paternalistic manner. According to this line of thinking, sometimes a putatively self-harming act is, in fact, a matter of ‘harm to others’. Yet, in this paper I argue that if we, as we ought to do, take into consideration the non-identity problem, this challenges the notion that the agent at T1 is in fact (...) imposing harm on anyone, even when we accept that he or she is numerically different from the agent at T2. If the life of the agent at T2 is still worth living, the agent is not worse off than he or she would have been in spite of the consequences pertaining to the putatively self-harming act since he or she would otherwise never have come into existence. In this way, the argument I put forward in this paper calls in question the ability of the shifting identity argument to actually justify imposing restrictions on self-harming behaviour. (shrink)