The application of advanced Artificial Intelligence techniques in astroparticle experiments represents a groundbreaking advancement in data analysis and experimental design. As space missions become increasingly complex, integrating AI technologies is essential for optimizing performance and enhancing scientific outcomes. This study focuses on the use of graph neural networks (GNNs) for tracking systems. Graph Neural Network are well-suited for tracking systems, as they can efficiently model the graph structure of detectors, where hits represent energy deposits and edges capture their relationships. We will present a novel approach that employs GNNs for node-level tasks, specifically aimed at classifying noise hits versus signal hits and reconstructing tracks.