In partnership with
Amazon has quietly shelved a three-year AI hiring project after internal tests revealed that the system consistently produced gender-biased results, according to reporting from Reuters.
The algorithm, designed to streamline recruitment by automatically ranking job applicants, showed a troubling pattern: résumés containing references such as “women’s”—for example, “women’s coding club” or “women’s leadership organization”—were rated lower. The tool also penalized graduates from several women-only colleges, highlighting a systemic skew in the model’s decision-making.
At its core, the flaw stemmed from the system’s training data. The AI had been fed résumés from a decade of Amazon applications—a pool largely dominated by men, particularly in technical roles. Instead of learning objective hiring criteria, the model absorbed and amplified existing workforce imbalances, incorrectly inferring that male applicants were inherently stronger candidates.
An Amazon insider described the team’s original ambition as a kind of automated shortlisting engine capable of sifting through hundreds of applications and surfacing the top contenders instantly. But despite extensive attempts by engineers to correct the bias, the model continued to behave unpredictably. By early 2017, leadership decided to discontinue the initiative.
This episode underscores a growing challenge for organizations embracing AI in talent acquisition. A CareerBuilder survey from 2017 found that over half of U.S. HR leaders expected AI to become a routine part of their work within five years—yet this case shows how easily automation can reproduce inequities if not carefully supervised.
Fairness, transparency, and explainability remain complex obstacles. As Carnegie Mellon University’s Nihar Shah noted, developing hiring algorithms that can be fully trusted and clearly understood is still a work in progress. LinkedIn Talent Solutions executive John Jersin echoed the sentiment, emphasizing that AI cannot replace the nuance and judgment of experienced recruiters.
Amazon’s abandoned tool now serves as a prominent reminder: without deliberate safeguards, automated systems may unintentionally reinforce the very biases organizations aim to eliminate.
Shoppers are adding to cart for the holidays
Over the next year, Roku predicts that 100% of the streaming audience will see ads. For growth marketers in 2026, CTV will remain an important “safe space” as AI creates widespread disruption in the search and social channels. Plus, easier access to self-serve CTV ad buying tools and targeting options will lead to a surge in locally-targeted streaming campaigns.
Read our guide to find out why growth marketers should make sure CTV is part of their 2026 media mix.
Stay updated with the latest insights—subscribe to our newsletter for real-time market analysis!