<p>This systematic review synthesizes 80 peer-reviewed studies published between 2000 and 2025 to examine artificial intelligence applications in visual and non-visual arts, based on PRISMA-guided searches across Google Scholar, Scopus, China National Knowledge Infrastructure, IEEE Xplore, and arXiv. Analyzing detection, recognition, interpretation, and generation domains across feature-based methods, convolutional neural networks, transformers, and diffusion models applied to Western and non-Western traditions including Chinese calligraphy, Indian ragas, African art, and Pre-Hispanic Latin American art—the review identifies three critical systemic gaps: non-Western artistic traditions appear in only approximately 11% of studies, evaluation methodologies lack standardization across domains, and prevailing ethical frameworks predominantly reflect Western institutional perspectives. These limitations introduce risks of cultural misrepresentation, appropriation, and authorship ambiguity when AI systems trained on Western-centric datasets are applied to Indigenous and non-Western art forms without community-based consent. To address these gaps, the study introduces a unified taxonomy, a historical chronology of AI–art development, and an ethical framework with defined validation pathways structured around four pillars—transparent dataset documentation, artist and community consent protocols, cultural sensitivity audits, and explicit human–AI attribution mechanisms. The validation pathways include stakeholder interviews, quantitative bias audits, cross-cultural surveys, and interdisciplinary oversight. By operationalizing established policy foundations, including UNESCO’s AI Ethics Recommendations, PRISMA standards, and Fair-ML principles, this review provides a consolidated foundation for responsible, cross-cultural, and ethically grounded AI research in creative domains.</p>