Hate speech is referred to as a form of communication that directly addresses insult or discrimination against individuals or groups. It may include discrimination based on race, religion, gender, sexual orientation, disability, or nationality. Over the internet, where we use a lot of social media, the identification of these hate commuters is challenging due to the nature of these platforms, which provide a hidden self-identity. The only possibility of stopping these people is to restrict and eliminate the streaming of objectionable content before it becomes widespread over the internet, thereby preventing chaos and further social damage. In this paper, we will discuss techniques adopted to create a Political hate-check model on political tweets collected by retrieving tweets from the Twitter platform related to the farmer protests in India and annotating the Tweets to train an LLM known as BERT, for identifying hate speech. The research focused on using of a pre-trained text model and then optimising fine-tuning methods to reach state-of-the-art accuracy. The model performed at an accuracy close to 91%. We also performed a comparative analysis and compared the model with other existing methods to establish that BERT performed far better than the other conventional models.