Deep generative models, particularly Generative Adversarial Networks (GANs), have become a popular tool for synthetically generating Electronic Health Records (EHRs) to mitigate the privacy concerns of sharing and using such data. Motivated by the performance of diffusion models over GANs on generating other data modalities such as image, text and sound, we propose medDiffusion, a new approach based on diffusion models to generate realistic tabular EHRs. In this talk, I will motivate our work, briefly introduce diffusion models and discuss our preliminary experimental results.