A federal judge just declined to throw out the AI discrimination case against Workday. Again.
If you’ve been following Mobley v. Workday, you know this case has been quietly rewriting the rules for every organization that lets software make the first cut. If you haven’t, here’s the catch-up:
→ 2023: Derek Mobley sues. He’s Black, over 40, and disabled. He was rejected from more than 100 jobs screened by Workday’s AI.
→ 2024: The court rules Workday can be held liable as an “agent” of the employers using its tools. The vendor is not a bystander.
→ 2025: The case is certified as a nationwide collective action. Workday’s own filings acknowledge that 1.1 billion applications were rejected through its system.
→ June 22, 2026: The judge lets the age and disability claims move forward, including one over the use of “proxy health indicators” and dismisses the race-based claims this round.
Read that last line again.
The part most coverage skipped
Same plaintiff. Same tool. Same 100-plus rejections.
The age claim survived. The disability claim survived. The race claim got dismissed.
That contrast is the story, and almost nobody covering this ruling led with it.
It tracks with something practitioners already know but rarely say out loud: algorithmic race discrimination is the hardest claim to keep alive in court. Age has a date on an application. Disability has an accommodation request, a gap, a proxy health indicator someone can point to. Race hides in variables few will name and in training data that learned from who did and who didn’t get hired before.
It’s a pattern, and it’s my read on why these claims break the way they do.
Dismissed is not the same as didn’t happen.
What this means if you’re job searching
That pile of silent rejections may not have been about you.
A machine made a human decision it had no business making and the age and disability claims are now being tested in open court. Not in a think piece. Not in a LinkedIn comment section. In front of a federal judge, with a nationwide class behind it.
You were not imagining the pattern. You were describing it before the courts caught up.
What this means if you’re hiring
Your vendor’s bias becomes your liability.
The 2024 agent ruling settled that question. “The algorithm did it” is not a defense, and “our vendor handles that” is not a compliance strategy. If a tool is making cuts on your behalf, you own the outcome of those cuts.
And here’s the uncomfortable follow-on: the claim that’s hardest to litigate is the exact one you should be auditing for hardest. Race claims fail in court not because the discrimination isn’t there, but because it’s buried in proxies. The fact that a plaintiff struggles to prove it says nothing about whether it’s happening inside your funnel.
If you can’t answer what your screening tool weighs, what it was trained on, and who it filters out, you don’t have a hiring process. You have a liability with a dashboard.
Where to start
I built a free Ethical AI Screening Scorecard for exactly this moment — nine questions to pressure-test the tools making your cuts.
Employers who want more than a self-check: I run full screening audits: vendor documentation review, disparate impact analysis, and a remediation roadmap your legal team can actually use. Let’s talk.
I’m not against AI. I’m against using it to make human decisions.
