Model Risk arises when the model's assumptions do not exactly correspond to reality. For example, the Black & Scholes model is based on log-normal distribution. Such assumption eliminates the possibility of a Market crash almost completely, making it necessary to adjust the model for the probabilities of extreme Market developments.

(See Chapter 24 Hybrid securities of the Vernimmen)

To know more about it, look at what we have already written on this subject