White men applying for executive positions have a new hurdle in their way: artificial intelligence.
Fresh research promoted by a new AI watchdog has revealed that resume audit filters used by mega-corporations to help identify the best candidates for top jobs have a secret bias that favors black people over white people and women over men.
Recommended Stories
- Art of the deal takes center stage: White House Report Card
- Vance approval matches Trump’s, giving MAGA a one-two punch
- Gun groups push for ‘generational win’ to end taxes, registration
Overall, the bias favors black people and women by 12%, but in some tests of the “cultural” words or references used in resumes between black women and white men, such as the colleges they attended, the bias can reach 50%.
“Given that these systems are already being deployed at scale in high-stakes decision-making scenarios, this represents an urgent problem requiring immediate attention,” said the scholarly study of AI resume audits used by major employers to pick candidates for attention and eventually a live job interview.
The surprise finding: Potential employers do not want any race or sex filters, and the AI companies are not doing enough to control for the bias.
“There’s up to a 12% difference in interview rates between white men and black women, despite, you know, all having identical resumes,” said Jason Hausenloy, an independent AI safety expert who is also associated with the newly established Washington watchdog Alliance for Secure AI.
Earlier this week, the Alliance for Secure AI officially launched as America’s newest AI safety group—united by a clear mission to ensure AI advances in a way that is safe and secure for humanity.
— The Alliance for Secure AI (@secureainow) June 6, 2025
The era of AI security is stronger than ever. #AI pic.twitter.com/hZdNbM4Syt
He noted that in one test cited by the research paper, exact resumes were fed through a so-called “large language model” used by companies to whittle down applicants. The name on one was “Tamika Williams (she/her),” a black candidate. The white candidate was referred to as “Todd Baker (he/him).”
Hausenloy said that the AI filter favored “Tamika Williams” 25%-50% more.
In another test of AI resume audits, colleges were used to indicate race. For black people, the study chose Howard University and Morehouse College, two Historically Black Colleges and Universities. To signal white candidates, Georgetown University and Emory University were used. Again, the black resumes won out.
“In some sense, it’s classic DEI,” said Hausenloy. But, he added, companies using the commercial resume filters are aware of the bias and aren’t demanding diversity, equity, and inclusion considerations.
SEE THE LATEST POLITICAL NEWS AND BUZZ FROM WASHINGTON SECRETS
The study, shared with Secrets, said that in the push for hiring efficiency, good candidates could be left behind if they don’t fit AI’s bias favoring black people and women.
“Proponents claim these LLM-driven assessments offer unprecedented efficiency and objectivity for tasks like resume screening and candidate interviewing. However, this rapid, widespread adoption raises significant concerns about fairness and bias,” said the study.