Research on open-source large language models (LLMs) revealed significant racial and gender biases in resume screening. The models strongly favoured white and male candidates, with Black men facing the most discrimination. These biases likely stem from societal privileges reflected in the LLMs' training data.
