The emergence of digital news sites provides us with information instantaneously across the globe and has shifted the way material is written and consumed. Nevertheless, it has hastened the spread of biased reporting, often supporting ideological echo chambers and societal polarization. Determining political bias manually at scale is subjective and can’t be done practically. This paper discusses an automated deep learning-based system for detecting and assessing political bias in news articles. The model will classify news articles into left, center, or right ideological categories, and indicate if there is any bias and its level, if there are any. The system has multiple components, including text pre-processing, product scraping of live data, and two main deep learning models. We will use a fine-tuned LLaMA 3.2-3B model to classify the bias and a fine-tuned MPNET model to classify semantic similarity to categorize articles associated with the same data event. Overall, this approach will significantly improve upon previous manual and rule-based systems by giving automated and real-time and scalable analysis with sub-contextual insight.