A Systematic Review of Publicly Available AI Ethics Resources for Media Production: Distribution, Documentation, and Research Gaps

This systematic review project identifies, analyzes, and synthesizes 50 publicly available AI ethics evaluation resources for media production released between 2020 and 2026. Following PRISMA 2020 guidelines, a structured search across nine academic platforms (arXiv, Google Scholar, Scopus, Hugging Face, Zenodo, Figshare, GitHub, Papers with Code, and OpenReview) was conducted to locate datasets, benchmarks, and evaluation frameworks explicitly designed to assess bias, authenticity, fairness, governance, and cultural diversity in AI systems used for news, social media, advertising, and video production. The resulting curated inventory documents each resource's ethical category, media domain, geographic focus, license type, documentation quality, and persistent identifier. The analysis reveals a predominance of bias detection datasets (28%), a heavy Western-centric focus (81%), and significant gaps in advertising (4%) and film (0%) domains. While documentation practices show improvement (78% with annotation protocols), only 54% use open licenses and 36% include formal datasheets. The project contributes a comprehensive, machine-readable metadata dataset under a CC0 license) that enables researchers, practitioners, funders, and regulators to identify existing tools, recognize critical gaps, and develop more inclusive, transparent, and practically useful benchmarking resources for ethical AI in media production.