We focus on the modeling of the survival function based on some covariates under the assumption of dependent censoring. In a lot of survival analysis, failure time is assumed to be independent with censoring time, because its dependency is not identifiable without additional information. Therefore, a sensitivity analysis for assessing the changes of estimates by the dependency is important but the operation is slightly annoying and there is the difficulty to construct the stable model finally. To address this problem, we propose a construction method of a survival tree under the dependency between failure and censoring. To construct the model, we assume that the subjects included in some node have the constant risks of the event and censoring. In addition to this, we assume that the joint distribution of the failure and censoring which are given by a copula with an unknown parameter. Then, we can estimate the parameters in the model by EM algorithm. Using the estimated parameters, the node is splitting by the exponential log-likelihood measure. We study the performance of this method by simulation studies. Additionally, we show the results of applying the method to actual data.