The project “6G Trans-Continental Edge Learning” (6G-XCEL) seeks to develop a framework for decentralized multi-party, multi-network artificial intelligence (DMMAI) in future 6G networks. 6G-XCEL will broadly engage with 6G stakeholders across the EU and the US through EU-US partnership in creating the DMMAI framework. Artificial Intelligence (AI) is increasingly recognized as a fundamental pillar for future 6G networks, enabling autonomous, intelligent, and adaptive network operations. Existing research has made significant progress in leveraging AI to optimize network performance; however, current studies largely focus on isolated AI controllers within specific domains, such as Radio Access Networks (RAN) or Core Networks (CN). There is limited research on the coordination and interaction of multiple AI controllers across diverse network environments. Furthermore, edge networks introduce additional complexities such as the diversity of computing platforms, the aggregation of multiple stakeholders, and the constrained and dynamic nature of such networks. The objective of the DMMAI framework is to close the gap in research and design by enabling the federation of AI-based network controls across network domains and layers. The goal of this deliverable is to describe the proposed architecture of the DMMAI framework. To do that, this report outlines: a definition of the framework, along with its objectives, the current state of the 6G architecture, and how the DMMAI will sit on it, a thorough description of the different parts of the framework, a list of security threats that will occur from it, and the implementation approach followed by 6G-XCEL team up until M16 of the project
6G-XCEL Consortium (Fri,) studied this question.