Embedding AI ethics in the data lifecycle: A framework for enterprise AI governance

Enterprises adopting AI often face a persistent gap between ethical principles and concrete operational practices. To address this challenge, this study proposes a data-centered governance framework that embeds ethical considerations throughout the AI data lifecycle. Using a mixed-methods approach, we analyze internationally representative AI ethics and data governance policy documents through grounded theory, informed by Value Sensitive Design and socio-technical systems theory. The resulting framework comprises four governance dimensions—macro values, technical foundations, actor practices, and organizational arrangements—operationalized through twenty indicators. Indicator weights are further derived using the Fuzzy Analytic Hierarchy Process based on assessments from enterprise practitioners. The findings reveal a strong emphasis on institutional accountability and technical compliance, while commitments to value-level ethical ideals and internal governance capacity receive comparatively less attention. Building on these results, we propose a practical governance framework that integrates accountability mechanisms, technical safeguards, and normative principles across the data lifecycle. Supplementary qualitative insights from cross-sector interviews further illuminate sectoral differences in implementation. Overall, the framework supports enterprises in operationalizing AI ethics and offers policymakers an empirical basis for bridging the ethics–practice gap.