Privacy‐Preserving Decentralized <scp>AI</scp> for Secure Data Sharing in <scp>6G</scp>

ABSTRACT Large‐scale cross‐domain data exchange via 6G networks is needed for intelligent applications. Centralized AI frameworks expose sensitive data, compromising privacy and security. This study proposes a privacy‐preserving decentralized artificial intelligence (PPDAI) system for safe and verifiable data transmission using federated learning (FL) and blockchain‐based access control (BAC). Avoiding data breaches and unwanted access, PPDAI saves raw data locally and delivers encrypted model updates via an immutable blockchain ledger. Breakthroughs include a hybrid privacy‐preserving method that combines model aggregation with decentralized ledger verification and a lightweight consensus mechanism for communication efficiency. Testing shows that PPDAI increases model accuracy by 4%–6%, decreases communication cost, and minimizes inference attack susceptibility compared to standard FL systems. The strong, scalable, and privacy‐aware basis for AI deployment in next‐generation 6G networks advances safe data sharing and decentralized intelligence frameworks.