Stanford HAI released a study that maps how hiring AI filters bias Black and Asian applicants. Researchers tracked 3.4 million job seekers who sent 4 million applications to 1,700 postings across 150 firms. Every application passed through a single vendor’s machine‑learning model before employers received a “recommend” or “reject” tag.

The analysis applied EEOC’s four‑fifths rule and found that 26 % of Black and 15 % of Asian candidates faced discrimination in at least one position. If the tools had treated all races equally, roughly 40,000 more applications would have moved forward, widening the talent pipeline for underrepresented groups.

A second finding shows that algorithmic monoculture drives systemic rejection. When applicants applied to four jobs screened by the same vendor, ten percent were rejected from every position—an outcome far higher than the baseline of independent company decisions. This pattern underscores how market concentration can lock out qualified candidates.

The report warns that widespread adoption of opaque AI tools—highly consequential and entrenched—compounds bias and denies talent. Without independent audits, employers risk perpetuating inequity and misallocating workforce potential. The findings call for evidence‑based policy to regulate AI hiring practices.