Summary: We propose a new estimator of the probability density function when the data is randomly censored, obtained through an EM algorithm, for solving a “nonlinearly smoothed” maximum likelihood problem. This algorithm is based on setting the Gateaux derivatives, in all directions, of the smoothed likelihood function equal to zero. Simulation results are presented which suggest that the proposed estimator compares favorably to the Kaplan-Meier-based kernel density estimator.