Explainable artificial intelligence (XAI) is an emergent requirement for the responsible use of AI applications in sensitive application domains due to technological advancements, regulatory pressures and ethical obligations, and alignment with the United Nations Sustainable Development Goals (SDGs). In this paper, we provide a survey of literature from 2023 to 2025, synthesizing the latest advances in approaches, performance metrics, applications and policies, with explicit emphasis on how XAI advances SDG 3 (Good Health and Well-being), SDG 9 (Industry, Innovation and Infrastructure), SDG 10 (Reduced Inequalities), and SDG 16 (Peace, Justice and Strong Institutions). We propose an enhanced classification framework that distinguishes between intrinsic versus post-hoc approaches, local versus global explainability, and emerging paradigms (multimodal, LLM-oriented, and interactive explanations) that extend classical taxonomic dimensions as orthogonal descriptors. We also present technical versus human-centred metrics, emphasizing fidelity, robustness, complexity, impact on decisions and evaluability, as well as shedding light on human-centric aspects. The survey investigates key challenges in scalability, security, privacy, ethical trade-offs, and their domain-specific applications in healthcare, cybersecurity, finance, and law. To bridge the research-practice gap, we present a practical “XAI-by-design” engineering template. Although XAI techniques are now mature, challenges related to reproducible research, interactive explanation, and dynamic regulatory adaptation remain significant obstacles. We conclude with concrete research advice to explore future research opportunities.
Pujari et al. (2026) studied this question.