Background: The protein p53 has a well established role in protecting genomic integrity in human cells. When DNA is damaged p53 induces the cell cycle arrest to prevent the transmission of the damage to cell progeny, triggers the production of proteins for DNA repair and ultimately calls for apoptosis. In particular, the p53-Mdm2 feedback loop seems to be the key circuit in this response of cells to damage. For many years, based on measurements over populations of cells it was believed that the p53-Mdm2 feedback loop was the responsible for the existence of damped oscillations in the levels of p53 and Mdm2 after DNA damage. However, recent measurements in individual human cells have shown that p53 and its regulator Mdm2 develop sustained oscillations over long periods of time even in the absence of stress. These results have attracted a lot of interest, first because they open a new experimental framework to study the p53 and its interactions and second because they challenge years of mathematical models with new and accurate data on single cells. Inspired by these experiments standard models of the p53-Mdm2 circuit were modified introducing ad-hoc some biologically motivated noise that becomes responsible for the stability of the oscillations. Here, we follow an alternative approach proposing that the noise that stabilizes the fluctuations is the intrinsic noise due to the finite nature of the populations of p53 and Mdm2 in a single cell. Results: We study three stochastic models of the p53-Mdm2 circuit. The models capture the response of the p53-Mdm2 circuit in its basal state, in the presence of DNA damage, and under oncogenic signals. They are studied using Gillespie's simulations, mean field methods and analytical predictions within the context of the Linear Noise Approximation. For the first two models our results compare quantitatively well with existing experimental data in single cells. The study of the response of the p53-Mdm2 circuit under oncogenic signals is process for which we do not have single cell measurements, but can be modeled waiting for experimental verifiable results. While we can not discard that other sources of noise in the cells may also be important, our results strongly support the relevance of the intrinsic noise in these systems. Conclusions: We suggest that the noise induced by the finite size of the populations is responsible for the existence of sustained oscillations in the response of the p53-Mdm2 circuit. This noise alone can explain most of the experimental results obtained studying the dynamics of the p53-Mdm2 circuit in individual cells.