This study investigates the potential biases of large-scale language models (LLMs), which play an increasingly important role in information dissemination and decision-making processes, particularly in the political realm. We assess the political bias and stereotype propagation of eight major LLMs using the two-dimensional Political Compass Test (PCT). We assess the inherent political leanings of the models and explore explicit stereotypes across various social dimensions using persona prompting using the PCT. Finally, we uncover implicit stereotypes using a multilingual version of the PCT.
