Abstract
The rapid expansion of Machine Learning (ML) across finance, healthcare, education, and public policy makes ethical oversight an imperative rather than an optional add-on. This paper responds to that urgency by proposing a comprehensive framework grounded in ten principles—accuracy, fairness, accessibility, security, privacy, transparency, accountability, human oversight, sustainability, and harm avoidance—and positioning them within existing international guidelines. Recent scoping reviews have highlighted the lack of consistent evaluation frameworks across domains and have called for systematic approaches to fairness, accountability, transparency, and ethics. Motivated by case studies of algorithmic redlining, dataset bias, hallucinations in large language models, and ecological concerns, we develop a weighted scoring rubric with thresholds to diagnose ethical compliance. We demonstrate the rubric through case studies, illustrating how the scores identify deficiencies and guide mitigation. The proposed framework is built upon the EU AI Act, NIST’s AI Risk Management Framework, UNESCO’s recommendations, and the OECD AI Principles. We reflect on AI’s energy footprint and the so-called “nuclear dependence” argument, and conclude with a roadmap for practitioners.