Workday hiring-AI class action advances in federal court

The interesting legal thread this month is not another frontier-model lawsuit, it is a hiring case that keeps quietly getting bigger. In Mobley v. Workday, plaintiff Derek Mobley applied to over 100 positions using Workday's screening software and was rejected by all of them. On June 22 a federal judge in San Francisco ruled his case could proceed, as Tech Policy Press lays out in an essay by Samonne Montgomery, who holds a PhD in Planning and Public Policy from Rutgers. His theory is that the vendor's tools breach the Americans with Disabilities Act and California civil-rights law through proxies for disability, including employment gaps.

The reason to care is a doctrinal one. Montgomery's argument is that institutions have learned to route uncomfortable decisions through software at the precise moment that assigning them to a person would cost the institution something. She borrows the "moral crumple zone" idea from researcher Madeleine Clare Elish to describe the human reviewer positioned near the decision but without authority to reverse it, what she calls "agency without authority." Courts are starting to reject the drape. She points to a 2024 British Columbia tribunal that held Air Canada answerable for the bereavement-fare misinformation delivered by its customer-service chatbot, refusing the airline's framing that the bot was "a separate entity." Her line for the hiring context is blunter: "An employer answers for a discriminatory outcome whether it built the screening tool or bought it."

The stakes for anyone running vendor AI in a hiring pipeline are direct. If Mobley proceeds to a class, the discovery target is both the buyer and the vendor, and "we outsourced it" starts looking thin as a defence. The honest caveat is that this is an essay about how responsibility gets assigned, not a prediction of who wins on the merits. The regulatory backdrop is also more permissive than the headline suggests: on June 16 the European Parliament postponed AI Act obligations for high-risk hiring systems until December 2027, and Rep. Nathaniel Moran's AI Incident Reporting Act, introduced June 25, focuses on frontier models rather than hiring software.

What the reporting does not give you is the size of the potential class, the damages exposure Workday actually faces, or whether the state-level bias-audit patchwork can fill the federal gap in the meantime. The direction that a leader should watch is the one that keeps recurring in these cases: in Montgomery's phrase, a human who can only approve is evidence of process, not accountability. If your compliance story rests on that reviewer, it is worth checking this quarter whether they can actually say no.