The ability to control for the kinds of information encoded in neural\nrepresentation has a variety of use cases, especially in light of the challenge\nof interpreting these models. We present Iterative Null-space Projection\n(INLP), a novel method for removing information from neural representations.\nOur method is based on repeated training of linear classifiers that predict a\ncertain property we aim to remove, followed by projection of the\nrepresentations on their null-space. By doing so, the classifiers become\noblivious to that target property, making it hard to linearly separate the data\naccording to it. While applicable for multiple uses, we evaluate our method on\nbias and fairness use-cases, and show that our method is able to mitigate bias\nin word embeddings, as well as to increase fairness in a setting of multi-class\nclassification.\n