Atypicality is a new concept that uses a codelength-based deviation from the norm to find the interesting rare events. In a previous paper we have developed an information theoretic approach for discrete data. Then in the two other papers, we came up with an extension to the real-valued models for Gaussian and vector Gaussian cases. In the current paper we generalize our real-valued model to the class of exponential family. We include a comprehensive table that summarize the approximated atypicality criterion and bounds for common distributions in the exponential family model.