One good solution has materialized in terms of adapting machine learning models to changing data privacy and fairness requirements-unlearning using machines. This paper explores an implementation of unlearning using neural networks on the Fashion MNIST and CIFAR datasets. We present a method to remove data selectively without having to retrain the full model, promising efficiency as well as legal compliance with GDPR. We illustrate the impact of unlearning on model accuracy and loss to determine what balance might be struck through unlearning between performance and privacy. Our results underline the strength of this notion in realizing the desired notion of ethical, secure, adaptive machine learning systems.