This study investigates the growing dependence of students on popular generative artificial intelligence tools such as ChatGPT, GitHub Copilot, Jupyter AI, Google Bard, and more for coding assignments, project development, exam preparation, and conceptual learning in computer science higher education. Based on empirical data from various undergraduate and graduate-level programming courses at the Academy of Applied Technical and Preschool Studies in Serbia, the research applies natural language processing (NLP) and builds a machine learning model to examine student engagement and predict potential educational outcomes. A Python-based comprehensive framework integrates CodeBERT method for semantic similarity and plagiarism detection, TF-IDF with cosine similarity for benchmark comparisons, XGBoost for rubric-based classification, and DBSCAN clustering methods for code anomaly detection. Sentiment analysis further captures student attitudes toward frequent AI use. Rather than limiting the use of such approaches, this paper introduces a scalable solution for AI-aware assessment and curriculum design, encouraging responsible and ethical usage of modern generative technologies. The results support an innovative and future-ready model for education in the era of artificial intelligence.