Central Bank Digital Currencies (CBDCs) boost digital payment system stability while increasing both efficiency and accessibility because they support direct money transfers between financial entities, which include both institutional and private institutions. The implementation of Transnational Confidentiality measures faces difficulties because it must fulfill both privacy requirements and the regulatory need for AML and CFT compliance. Our solution consists of Agentic AI-driven CBDC Systems, which use privacy- enhancing technologies (PETs) and AI-based compliance systems to solve this issue. Our system combines zero-knowledge proofs (ZKPs), homomorphic encryption (HE), blind signatures, secure multi-party computation (MPC), and trusted execution environments (TEEs) for performing confidential transactions that remain auditable. The system adopts a dual UTXO with account mechanisms to achieve transaction privacy, and it provides regulatory transparency that relies on controlled disclosure controls. Our solution merges FL and ZKML to perform decentralized anomaly recognition and fraud threat evaluation without revealing crucial financial information. Testing and refinement were performed in multiple rounds with specialists in DLT and cryptography, as well as financial regulations until the system met practical requirements and regulatory standards. When Agentic AI operates with advanced cryptographic primitives, it establishes a secure framework for next-generation CBDC ecosystems that maintains scalability and regulatory compliance. The presented research develops auditable CBDC systems that maintain privacy protection, thus leading to the development of secure financial infrastructure.