News recommendation systems (NRS) are increasingly shaping online news consumption and thus the world perception of many readers. In NRS, state-of-the-art deep learning models are becoming increasingly popular. In these models, unwanted biases in the training data can lead to bias in the model. In recent years, there has been a surge of research on bias in recommendation systems examining bias in word, text, and knowledge graph embeddings. Literature on analyzing biases in research models proposed for news recommendation is still scarce. To translate scientific research models into responsible NRS bias analysis is indispensable. This thesis investigates bias in the KRED NRS, a content-based knowledge-aware NRS that uses a knowledge graph as side information. To analyze KRED’s recommendations regarding political bias in exposure, we create a corpus of news articles on the politically divisive topic of migration in the EU. Using this article corpus, six news reception profiles, and synthetically generated user behavior, we analyze the diversity of recommended articles. The analysis shows that the KRED NRS preferably recommends news from news outlets more frequently contained in the user‘s reading history. ... mehrIndependent from the user‘s news reception profile, the KRED NRS unfairly favors news articles from Breitbart (right-wing political bias) in the article corpus.