The rising complexity and frequency of cyber threats necessitate advanced approaches to detect and mitigate intrusions in information systems. This research explores the use of transformerbased models for anomaly detection in cyber security, leveraging Hugging Face’s state-of-the-art tools to improve the identification of malicious activities. The study focuses on analyzing wellknown datasets, such as NSL-KDD and CICIDS, to identify unusual patterns indicative of cyberintrusions. Various preprocessing steps are applied, including data normalization and feature selection, to optimize the dataset for model training. The transformer models are then implemented to classify normal and anomalous network behavior, aiming to achieve higher detection accuracy and reduce false positive rates. Performance is assessed using metrics like precision, recall, and F1-score, with baseline comparisons made against traditional detection methods. The results are expected to demonstrate the superior capability of transformer-based models in handling complex intrusion patterns. This research contributes to the development of more robust and reliable intrusion detection systems, providing insights into the application of advanced AI models in cyber security.