New research from the University of Washington found significant racial, gender and intersectional bias in AI large language models (LLMs) used for hiring processes. The researchers tested more than 550 real-world resumes with LLMs from Mistral AI, Salesforce and Contextual AI and found that these models favored resumes with white-associated names 85% of the time, male-associated names 52%, and rarely preferred Black male names.
The study highlighted how intersectionality, particularly for Black men, exacerbated bias, revealing that Black male names were consistently ranked the lowest, while Black female names were preferred more often than male names. With AI tools rapidly being integrated into hiring, the researchers stress the need for better regulations and research to ensure fairness and reduce harm, especially considering the increasing use of these systems by businesses.