Privacy risks in machine learning systems manifest in various ways, chief among them being the potential for unintended data exposure. These risks include the ability of an adversary to perform membership inference attacks, determining whether a specific individual's data was used in training a model, and model inversion attacks, where sensitive attributes about individuals can be reconstructed from model outputs. Additionally, machine learning models can unintentionally memorise and reveal parts of the training data. These risks highlight the delicate balance between leveraging the power of machine learning and protecting individual privacy, necessitating robust privacy- preserving techniques in model design and deployment.