Deliverable D4.1: Cognitive coordinator, AI agents and User-centric functions

This deliverable presents the architectural design and preliminary implementation of the SAFE-6G cognitive trust management framework. It focuses on the Cognitive Coordinator (CoCo), eXplainable AI (XAI) module, and the five user-centric Trust Functions (TFs): Safety, Security, Privacy, Resilience,and Reliability. The CoCo is implemented with modular architecture, including a BERT-based regression model to interpret user intent and a resource-aware reasoning engine to calibrate trust levels in real-time. This module orchestrates the actions of the TFs based on semantic inputs and system constraints. The XAI module, developed in parallel, was designed to provide both local (contextual) and global (modellevel) interpretability for decisions made by the Trust Functions.Major technical progress has been achieved across all TFs, with each now featuring a defined architecture, multi-layer interaction model (AI Agent, vApps, nApps), and functional logic for enforcement, monitoring, and adaptation.The Safety Function has delivered a modified Software Defined Perimeter (SDP) mechanism for secure resource isolation across the continuum.The Security Function integrates local AI classification and forecasting with verifiable credentials and a Hyperledger Fabric blockchain for access control and auditability.The Privacy Function enables fine-grained, session-level control using containerized nApps that modify 5G core behaviour according to dynamically computed privacy levels.The Resilience Function introduces AI-driven fault tolerance using rule-based orchestration and telemetry feedback across OpenCAPIF and aerOS interfaces.The Reliability Function maps Level of Trustworthiness (LoTw) tiers to tailored AI/ML pipelines and leverages multi-layer telemetry collection for Quality of Experience (QoE)-aligned prediction and mitigation. This deliverable marks a key milestone by establishing the foundation for SAFE-6G’s trust-aware orchestration layer. It prepares the ground for full-scale integration and validation activities, including the deployment of the CoCo, Trust Functions, and AI agents in immersive metaverse scenarios. These components will be integrated into advanced 6G experimentation platforms and tested under diverse threat models and trust configurations. These activities are coordinated as part of the project’s broader integration and pilot validation efforts, where system-level testing and use case execution will be carried out in the next phase and documented in the corresponding project deliverables.