The 6G networks are predicted to be based on artificial intelligence (AI) to support autonomous working, self-organization, and real-time decision making. Nevertheless, the extremely distributed, data-intensive, and heterogeneous characteristics of 6G settings pose significant problems with information privacy, data security, and node reliability. The extremely low latency requirements and the number of connected devices make traditional centralized security mechanisms no longer viable. In this paper, a new framework called Federated Trust-Aware Secure AI (FTS-AI) has been introduced to improve privacy, security, and resilience in the 6G system. The framework integrates federated learning (FL) to train models without centralizing raw data, blockchain to make model aggregation transparent and tamper-evident, and fuzzy logic-based trust management to continuously evaluate the reliability of nodes and identify malicious or compromised members. In the proposed design, 6G nodes locally train lightweight AI intrusion detection models and exchange only protected model updates during the collaborative learning process, never the raw data. A registry anchored on a blockchain holds verifiable trust scores and provides integrity along the aggregation pipeline. Empirical evaluation on the CIC-IDS2018 benchmark dataset shows that FTS-AI attains a detection accuracy of 96.3% and a false-positive rate of 3.2% under 20% malicious participation, corresponding to a relative reduction of 62% in false positives with respect to FedAvg, while converging in fewer communication rounds than state-of-the-art robust aggregation baselines.
