KHVPF Insight
Workday Discrimination and Generative Bias: Who is Responsible When AI Goes Bad?
Anyone who doubts the rapid integration of artificial intelligence into the workplace, official or unofficial, need only look to the events of June 10, 2025. A nearly day-long ChatGPT outage prompted more than 500,000 Google searches and thousands of social media posts from office workers so reliant on the technology they couldn’t imagine doing their jobs the old-fashioned way. It’s no surprise, then, that AI is now increasingly used for fundamental human resources functions, including decisions about who gets hired and who doesn’t. But this new technology has inherited old problems, chief among them the risk of discrimination in hiring.
Courts are now beginning to grapple with claims of algorithmic discrimination. One of the most high-profile cases is Mobley v. Workday, Inc., currently pending in a California federal court. Workday offers a subscription-based AI hiring platform that it promotes as “reducing time to hire by automatically dispositioning or moving candidates forward in the recruiting process.” Derek Mobley—a Black man over 40 with disabilities—allegedly applied more than 100 jobs through companies using Workday’s system and was rejected every time, sometimes within an hour. He claimed that Workday’s algorithm, functioning as an agent of its client employers, has a disparate impact on job seekers who are over 40, have disabilities, or are Black.
The court found that Mobley’s allegations were sufficient to state a legal claim. It accepted the argument that Workday’s AI could be considered an agent implementing its customers’ HR policies. And the Mobley court recently conditionally certified a collective action under the Age Discrimination in Employment Act for individuals over 40 who were rejected by Workday’s algorithm. According to Workday’s briefing, this group could include millions of people. Although Workday is the defendant here, there’s no reason that a rejected candidate couldn’t bring a lawsuit against the employer itself for using these tools.
Similar, if smaller, lawsuits are emerging elsewhere. A Massachusetts job applicant sued CVS in federal court after learning that the company secretly used HireVue, Inc. to perform an AI-based analysis of a video interview before rejecting him. CVS settled the suit late last year. A different lawsuit against HireVue is now pending in Colorado, filed by a deaf applicant who was denied employment after an AI analysis concluded, among other things, that she was not “practicing active listening” during a video interview.
The regulatory landscape remains fragmented. The Biden Administration had issued executive guidance requiring the development of AI safeguards, including policies “consistent with the advancement of equity and civil rights.” On January 21, 2025, the Trump Administration rescinded that guidance. The EEOC promptly removed related resources from its website, including its recommendations for responsible use of AI in hiring.
Still, some states are stepping in. As of October 2024, the Michigan Civil Rights Commission has adopted Guiding Principles for the Elimination and Prevention of AI Bias, which highlight the risk of algorithmic discrimination and calls for legislation to ensure transparency, human oversight, and fairness in the use of workplace AI.
As the Workday and CVS cases show, existing anti-discrimination laws are already being used to challenge AI-driven employment decisions. Employers should not assume that automating hiring insulates them from legal liability. AI systems are designed by humans, and they are only as reliable as the data, assumptions, and algorithms that shape them. Employers must audit these tools for disparate impacts, just like any other business process. And they need to be ready to defend their use of these tools in terms that judges and juries can understand. That’s not a task to hand off to a vendor or an IT team; it requires legal analysis and judgment. Otherwise, the only thing AI may streamline is the path to a costly verdict.