Data clustering plays a critical role in various domains, yet traditional clustering methods often struggle with noise and uncertainties inherent in real-world datasets. This paper introduces a novel safe semi-supervised fuzzy clustering approach based on Picture Fuzzy Sets (PFS), termed NPFS3FCM, designed to address these challenges. By integrating PFS, which incorporates positive, neutral, and refusal degrees, with a safe semi-supervised learning framework, the proposed method enhances clustering accuracy and robustness, particularly in noisy and ambiguous data environments. Comparative evaluations on multiple datasets reveal that NPFS3FCM outperforms comparing to others in terms of classification accuracy, clustering quality by DB index, and computational time. These results highlight NPFS3FCM as a reliable and efficient solution for real-world data clustering applications, effectively balancing performance and scalability.