91 researchers found significant racial, gender and intersectional bias in how three state-of-the-art large language models ranked resumes. The models favored white-associated names 85% of the time, female-associated names only 11% of the time, and never favored Black male-associated names over white male-associated names.






































