A recent position paper argues that fairness failures in generative models are due to an evaluation problem. The authors claim that current methods for evaluating fairness are not comparable across papers and do not provide actionable results for deployment decisions. They propose a standardized approach, called Fairness Cards, which makes evaluation choices explicit and enables reproducibility, comparability, and accountability. --- Why it matters: This matters because it highlights the need for more rigorous evaluation methods in generative models to ensure fairness. Currently, developers are left with ad-hoc bias checks that do not provide actionable results, making it difficult to deploy fair models in real-world applications. Source: https://arxiv.org/abs/2608.16974