Generally, diseases are primarily harmful to the individual herself; harm to others may or may not be a secondary effect of diseases (e.g., in case of infectious diseases). This is also true for mental disorders. However, both ICD-10 and DSM-5 contain two diagnoses which are primarily defined by behavior harmful to others, namely Pedophilic Disorder and Antisocial (or Dissocial) Personality Disorder (ASPD or DPD). Both diagnoses have severe conceptual problems in the light of general definitions of mental disorder, like the definition in DSM-5 or Wakefield's "harmful dysfunction" model. We argue that in the diagnoses of Pedophilic Disorder and ASPD the criterion of harm to the individual is substituted by the criterion of harm to others. Furthermore, the application of the criterion of dysfunction to these two diagnoses is problematic because both heavily depend on cultural and social norms. Therefore, these two diagnoses fall outside the general disease concept and even outside the general concept of mental disorders. We discuss whether diagnoses which primarily or exclusively ground on morally wrong, socially inacceptable, or criminal behavior should be eliminated from ICD and DSM. On the one side, if harming others is a sufficient criterion of a mental disorder, the "evil" is pathologized. On the other side, there are practical reasons for keeping these diagnoses: first for having an official research frame, second for organizing and financing treatment and prevention. We argue that the criteria set of Pedophilic Disorder should be reformulated in order to make it consistent with the general definition of mental disorder in DSM-5. This diagnosis should only be applicable to individuals that are distressed or impaired by it, but not solely based on behavior harmful to others. For ASPD, we conclude that the arguments for eliminating it from the diagnostic manuals overweigh the arguments for keeping it.