Research: Zero-touch network and Service Management (ZSM) is a key pillar in 6G networks. It allows the 6G management and orchestration framework to operate the networks without external (e.g., human) intervention. To effectively achieve ZSM, advanced network management procedures are required to detect and resolve anomalies within the 6G network autonomously, which usually requires Artificial Intelligence (AI) and Machine Learning (ML) models. However, relying solely on AI can raise concerns about trust due to their lack of explainability. Indeed, as these models are not explainable, it is difficult to understand and trust their decisions. To overcome this limitation, this paper introduces a novel pipeline for ensuring trustworthy ZSM in 6G networks by combining: (i) AI for detecting anomalies; (ii) eXplainable AI (XAI) to identify the root causes of anomalies using feature importance analysis; and (iii) Large Language Models (LLMs) to generate user-friendly explanations and suggest/apply corrective actions to resolve anomalies. A use case is presented using XGBoost as AI, SHAP as XAI, and Llama2 as LLM to address Service Level Agreement (SLA) latency violations within cloud-native 6G microservices. Evaluation results obtained through real experiments demonstrate the framework's efficiency in scaling cloud resources to prevent SLA violations while providing understandable explanations to users, thereby enhancing trust in the system.

Abstract-Industry: Zero-touch network and Service Management (ZSM) is key to achieving fully autonomous 6G networks. This paper presents a trusted approach that combines Artificial Intelligence (AI), eXplainable AI (XAI), and Large Language Models (LLMs) to provide ZSM insights and actions. Tested in realistic cloud-native 6G environments, it successfully prevents Service Level Agreement (SLA) violations in microservices applications.