Data Ethics and Responsible AI Use: A Scoping Review of Governance Principles, Implementation Challenges and Future Directions

Abstract

Artificial intelligence is increasingly embedded in education, healthcare, finance, public administration, scientific research and media systems. These applications create efficiency and innovation but intensify concerns about privacy, bias, opacity, accountability, consent, surveillance, intellectual property, misinformation and the concentration of technological power. Data ethics is central to this debate because AI systems inherit the legal, social and epistemic properties of the data on which they are trained, evaluated and deployed. This paper examines the relationship between data ethics and responsible AI use by synthesising recent literature on ethical principles, AI governance frameworks, data stewardship, implementation barriers and sector-specific risks, and proposes an integrated framework for developing, deploying and monitoring AI systems in ways that are lawful, transparent, accountable, inclusive and human-centred. A scoping review and evidence-mapping design was adopted. Note for readers: the evidence base was developed as a structured evidence map rather than a fully registered systematic review with dual independent database screening; a reproducible search rerun using Scopus or Web of Science is recommended before final journal publication, and this limitation is discussed fully in the Limitations section. Traceable scholarly articles, open books, standards, legal instruments and official policy documents published between 2016 and 2026 were identified through targeted searches of institutional portals, citation chaining and open scholarly platforms. Evidence was charted and synthesised thematically across six literature clusters. Twenty-eight high-relevance sources were retained for primary synthesis, supplemented by additional methodological and comparative sources. The evidence shows strong convergence around fairness, transparency, privacy, accountability, human oversight, safety and inclusiveness, but weaker agreement on implementation. The most persistent gap is not lack of principles; it is the difficulty of converting principles into lifecycle controls, institutional accountability, enforceable standards and measurable outcomes. Recent evidence from 2025 and 2026 corroborates this finding: even as AI regulation has expanded globally, the proportion of organisations with formally implemented AI governance frameworks remains critically low. Developing countries face additional constraints related to regulatory capacity, digital infrastructure, local datasets, talent, procurement dependence and unequal bargaining power with global technology providers. Responsible AI cannot be achieved by technical optimisation alone. It requires ethically sourced and well-governed data, risk-based governance, documentation, participatory oversight, continuous monitoring, enforceable accountability and context-sensitive capacity building. The proposed six-layer integrated framework ethical leadership, data ethics, model governance, deployment controls, stakeholder assurance and continuous learning provides a conceptual pathway for researchers, policymakers, developers and organisational leaders seeking trustworthy AI systems. Empirical validation of this framework across sectors and regional contexts is an identified priority for future research.