Fake news detection has become a critical task in today’s world of rapid information exchange. With the rise of social media and online news platforms, it has become easier for fake news to spread and influence people’s opinions and beliefs. Fake news can have serious consequences, ranging from influencing election outcomes to causing public panic and social unrest. This has led to a growing interest in developing automated techniques for detecting fake news. Our proposed methodology is a transformer-based model, specifically fine-tuned BERT with a linear classifier. Systematic testing was conducted using a fake news dataset in English language obtained primarily from Kaggle which was extended by collecting data from websites, yielding significant performance with an accuracy of 99%. Additionally, a comparative analysis was carried out with other models, namely the LSTM classifier and Bi-LSTM classifier, using various word embedding techniques. Initially, the model was developed to detect fake news in the English language. This work was later extended to Tamil, one of the most widely spoken languages in the world, with over 70 million speakers, primarily in India, Sri Lanka, and Southeast Asia. This extended work classifies fake and real news in the Tamil language using fine-tuned multilingual-BERT, achieving an accuracy of 97%.