Evaluating generative AI safety on real-world harm: A focused classification study using the RealHarm dataset. AIW. 2026;1(2). Accessed October 28, 2026. https://www.aiandwisdom.org/index.php/aiandwisdom/article/view/14