AI Hiring Tools Show Racial Bias Against Black, Asian Applicants

More than 90% of U.S. employers now use AI-powered screening tools from third-party vendors to manage job applications, and new research reveals these systems are producing racially biased outcomes at scale.

A study titled "Algorithmic Monocultures in Hiring" examined how AI vendors influence hiring decisions by analyzing data from 3.4 million real job applicants and submitting 4 million test applications to 156 employers across 11 market sectors. The research team from Stanford University, Chapman University, and Northeastern University documented what they call the first large-scale evidence of racial disparities in high-stakes hiring decisions driven by algorithmic tools.

The bias in the numbers

The study found that AI systems from a single vendor favored white applicants over Black and Asian candidates. Specifically, 26% of Black applicants and 15% of Asian applicants applied for positions where the AI exhibited bias against their racial group. Had these candidates advanced at the same rate as white applicants, an estimated 40,000 additional people would have reached the next hiring stage.

"What we found is adverse impact for Black and Asian applicants in particular," said Stanford researcher Rishi Bommasani, who led the study. "For some positions, the hiring AI tool was less likely to recommend Black and Asian applicants to certain positions."

The research also uncovered what the team calls "algorithmic monocultures"—when multiple employers use the same vendor's AI tool, applicants rejected by one company are likely rejected by all of them, far more often than chance would predict. This pattern would likely differ if companies made independent hiring decisions rather than relying on shared algorithmic systems.

Why it matters

As application volumes surge—employers now review nearly three times as many applications for entry-level positions as they did in 2022—companies increasingly depend on AI to manage the workload. But the Stanford findings suggest this efficiency comes at a significant cost to fairness. When a dominant vendor's biased algorithm spreads across industries, individual hiring discrimination becomes a systemic barrier affecting tens of thousands of qualified candidates. The research provides concrete evidence for policymakers and business leaders who must decide how to govern AI in high-stakes decisions where external scrutiny has been minimal.

Calls for oversight

Sarah Bana, assistant professor at Chapman University and study co-author, noted that employers feel they have little choice but to use algorithms given the volume of applications. Yet Bommasani cautioned against wider AI adoption in hiring without better safeguards.

Aalok Mehta, director of the Wadhwani AI Center at the Center for Strategic and International Studies, warned that without human review and better vendor management, these patterns will replicate across industries.

The findings were first reported by AfroTech and detailed in a Stanford research blog post. Bommasani emphasized the need to build an evidence base around hiring AI, noting that these tools are "high stakes and incredibly prevalent, yet there's very little external scrutiny."