Harm Laundering in GPT Models: Evidence That Gender Discrimination Is Transformed Rather Than Reduced Across Safety-Trained Generations
Safety evaluations for large language models rely on surface-form classifiers that report declining harm scores across model generations. We provide evidence that this methodology is systematically incomplete: explicit discriminatory content is transformed rather than removed. We call this \emph{harm laundering}. Analysing 450,000 gender-directed completions across 15 models sp
Reading in Plain English mode — toggle below to switch
Safety evaluations for large language models rely on surface-form classifiers that report declining harm scores across model generations. We provide evidence that this methodology is systematically incomplete: explicit discriminatory content is transformed rather than removed. We call this \emph{harm laundering}. Analysing 450,000 gender-directed completions across 15 models sp
This research highlights a critical flaw in current LLM safety evaluation methodologies, demonstrating that perceived reductions in harm may instead be transformations. For ML engineers, understanding 'harm laundering' is crucial for developing more robust and equitable AI systems, moving beyond surface-level safety metrics to address deeper biases.