GEPI report: AI amplifies gender bias in hiring, lending and healthcare

The Gender Equity Policy Institute's report How AI Fails Women reviews research showing automated systems reproduce and intensify gender discrimination. Screening models favored male-associated names in 51.9% of résumé comparisons versus 11.1% for female names, while GPT-4 rated cardiac stress testing highly important in 71% of male cases and 58% of female ones.

• Screening models favored male names in 51.9% of comparisons, female names in 11.1%
• GPT-4 rated cardiac testing highly important in 71% of male vs 58% of female cases
• Women's loans earned higher returns, yet models overestimated their default risk
• Dropping the gender field alone fails: names and histories still carry signals