Amazon's experimental recruiting tool (2014-2017)
Reuters reported in October 2018 that Amazon had built an experimental AI recruiting engine starting in 2014. It scored resumes from one to five stars, trained on patterns in resumes the company had received over the previous decade. Most of those resumes came from men.
By 2015 the team saw the problem: the system had taught itself that male candidates were preferable. It penalized resumes containing the word “women's,” as in “women's chess club captain,” and downgraded graduates of two all-women's colleges. Amazon edited the model to neutralize those specific terms, but engineers could not be sure the machine would not find new proxies. The project was scrapped.
Amazon's position, then and now: the tool was experimental and was never used by recruiters to evaluate candidates. No regulator found otherwise. The case still matters because it is the cleanest public demonstration of the core failure: a model trained on history will reproduce history, including the parts you would not defend.
EEOC v. iTutorGroup: programmed age cutoffs (settled 2023)
In September 2023 the US Equal Employment Opportunity Commission announced a $365,000 settlement with iTutorGroup. The EEOC's lawsuit alleged the company's tutor application software was programmed to automatically reject female applicants aged 55 or older and male applicants aged 60 or older. More than 200 qualified US-based applicants were allegedly rejected because of their age.
This one is different from Amazon's story. There was no learned proxy and no emergent bias. The discrimination was written into the software as an explicit rule. The settlement also brought non-monetary terms: anti-discrimination training, a new policy, injunctions against age and sex discrimination in hiring, and EEOC monitoring for at least five years.
The lesson for the game: not every biased outcome comes from a mysterious model. Sometimes the machine does exactly what it was told, and what it was told to do was, the EEOC alleged, an ADEA violation.
Mobley v. Workday: can the vendor be liable? (pending)
Derek Mobley applied to more than 100 jobs through employers using Workday's recruiting platform and says he was rejected every time, sometimes within minutes. He sued in 2023, alleging the screening tools discriminated based on race, age, and disability.
On June 22, 2026, a federal judge in California largely denied Workday's motion to dismiss, letting claims under California's Fair Employment and Housing Act and the federal Americans with Disabilities Act move forward. The court treated the vendor as a potential agent of the employers using its tools, which means “the vendor's algorithm did it” is not a defense that ends the case early.
Nothing has been proven. The case is in litigation and the discrimination is alleged, not adjudicated. But the procedural ruling already changed the conversation: vendors that build hiring tools may have to answer for them directly, and employers may not be able to outsource the liability along with the software.
