A judge has ruled that Workday must confront a lawsuit in California regarding alleged bias in AI hiring practices.

      A judge in San Francisco has allowed a proposed class action to move forward, marking it as the first to significantly challenge the algorithms used in AI screening software. A federal judge in the U.S. has determined that Workday must respond to claims that its AI-driven hiring software excluded job applicants from other companies in ways that purportedly violated California law and federal regulations against disability discrimination.

      This ruling from U.S. District Judge Rita Lin in San Francisco keeps alive a case that tests whether the companies behind algorithmic recruiting tools can be held accountable for the actions of those tools. Lin dismissed Workday’s argument that California's anti-discrimination laws should not apply to it when its software screens applicants located outside the state applying for jobs elsewhere, as reported by Reuters. This jurisdictional issue was crucial for Workday, as much of its defense relies on the assertion that it merely provides software and does not engage in hiring, either in California or elsewhere.

      The case, Mobley v. Workday, was initiated in 2023 by lead plaintiff Derek Mobley, who is over 40, African American, and has a disability. It is noted as the first of its kind aimed at the algorithmic decision-making involved in AI screening software, rather than targeting the actions of any specific employer using the tool. Mobley claims he was turned down for numerous jobs that were processed through Workday’s platform.

      However, Lin’s ruling was not entirely favorable for the plaintiffs. She dismissed allegations that the software discriminated against Asian American applicants but allowed unrelated claims to continue that it disadvantaged Black applicants, women, and individuals over 40. The remaining claims are still allegations that the plaintiffs must substantiate; the ruling merely permits these arguments to be presented, not affirming their truth.

      The case had already met an important threshold. The court had previously authorized it to proceed as a collective action for applicants aged 40 and above who were rejected through Workday’s platform from September 24, 2020, onwards, potentially encompassing a significant number of rejected candidates. The latest ruling expands the legal theories that can be explored during the trial.

      Workday categorically denies the allegations, stating that they are untrue, that its AI recruiting tools “do not make hiring decisions in California or anywhere else,” and that the technology “examines only job qualifications, not protected characteristics like race, age, or disability.” The company asserts that it rigorously evaluates its products under a Responsible AI program, a defense that, if successful, would clearly distinguish the vendor from the employers that configure and utilize its software.

      This distinction is currently at the heart of the dispute. If a screening tool generates biased results, the unresolved question is whether the company that developed it can be legally considered an agent of the employers who use it. The implications extend beyond a single company; algorithmic screening has been quietly integrated into corporate hiring practices, and a so-called AI-on-AI conflict in recruitment has drawn attention toward jobs lost rather than candidates filtered out before a human reviews their application.

      Many of the largest employers in the U.S. utilize various platforms to process applicants, often without revealing that the first selection was made by an algorithm rather than a recruiter. A determination that the vendor can be held liable would alert all of those companies using such tools. Concurrently, scrutiny regarding how AI systems manage sensitive personal data has intensified.

      Currently, the immediate impact of Lin’s ruling is narrow yet significant: Mobley v. Workday will proceed, and the discovery phase will begin to quantify the claims being made.