Why the May 29 Workday Ruling Should Change How HR Teams Pick AI Hiring Tools

What the Court Did Last Month

On May 29, 2026, Magistrate Judge Laurel Beeler of the Northern District of California issued a discovery order in Mobley v. Workday, Inc. resolving three disputes:

Workday's internal bias-testing data is protected by attorney-client privilege. Plaintiffs cannot compel it.
Workday is not required to produce its customers' applicant data, because Workday does not "control" that data under Rule 34.
Workday's own EEO-1 and OFCCP filings are ordered produced.

This is a discovery order in one case, not a substantive ruling on liability and not binding precedent on other courts. But it draws a line that matters for any employer using an AI screening tool, and points to a single practical conclusion: your choice of tool, and how much it automates, is the part of your risk you must account for when hiring.

The Case Behind the Order

The named plaintiff is Derek Mobley, an African American man over forty who alleged various disabilities under the ADA. Beginning in 2017, he applied to over one hundred positions through employers using Workday and was rejected from every one. Four additional plaintiffs over forty joined him, alleging hundreds of similar rejections.

The complaint frames the harm structurally. The plaintiffs allege that Workday's AI screening systems reproduce and amplify historical hiring disparities through "proxy variables"; these are features that correlate with race, age, or disability without being labeled as such. The legal theory advanced is that Workday acts as an "agent" of its customer-employers when its tools (e.g., AI Screeners) score, sort, or screen candidates, and so falls within Title VII, the ADEA, and the ADA.

The Workday Products at Issue

The order names the specific tools in dispute:

Candidate Skills Match — Workday Recruiting's algorithmic feature that scores how well an applicant's skills match a role.
Spotlight — a candidate review tool that scores an applicant's fit against job requirements. Acquired by Workday in 2024 when it bought HiredScore.
Fetch — a sourcing tool that surfaces individuals for open roles. Also from the HiredScore acquisition.

All three produce algorithmic outputs that an employer's recruiter then acts on. Workday's relationship with its customers is governed by a Master Subscription Agreement under which "Customer owns all right, title and interest to its Customer Content." That ownership clause is load-bearing for the customer-data ruling discussed in detail below.

What the Order Actually Held

Bias-testing data: privileged

Workday represented to the court that its bias testing of Candidate Skills Match was conducted "at the direction and under the guidance of its legal counsel for the purpose of rendering legal advice," with access restricted to those who needed it to give that advice. This ruling rejected Plaintiffs' arguments that the underlying data is factual and not privileged, that counsel was acting in a "business capacity" (i.e., not a legal advisor), and that Workday waived privilege by publicly claiming in its AI Fact Sheet that it performs bias testing.

In rejecting each argument, the Court applied the eight-part Ruehle test for attorney-client privilege, and held that "Workday has shown more than mere direction by its attorneys," and that:

Workday's invoking the mere existence of its bias testing outside of litigation is not enough to waive privilege.

Two qualifiers matter. First, the court emphasized that privilege "is strictly construed." Workday cleared a high bar; it does not follow that all vendor bias-testing programs would. Second, Workday is not claiming privilege over everything. The court noted: "there may be some information that is not privileged" and Workday has disclosed underlying data it is not asserting privilege over. The order addresses only the bias-testing data Workday is actively claiming. The Court did state, however, that the testing data may be relevant to the proceedings, but was still privileged.

Notably, Workday did not claim privilege over a separate 2023 bias audit of Spotlight performed by an external consultant under NYC Local Law 144. Testing structured to face a regulator is treated differently from testing structured through counsel.

Customer applicant data: outside Workday's control

Plaintiffs sought the applicant data sitting inside Workday's customers' systems, pointing to a clause in the Master Subscription Agreement allowing Workday to disclose customer information "to the extent required by Law." But that same clause lets the customer "seek injunctive relief to enjoin any breach or threatened breach."

The court held that this conditional permission is not "control" in the Federal Rule 34 sense:

the court cannot say that Workday has a legal right to obtain its customers' data on demand where the mechanism for production is a court order that the customer can seek an injunction to prevent.

Plaintiffs were told to keep working with the customer third parties they had already subpoenaed directly.

EEO-1 and OFCCP filings: ordered produced

The court rejected Workday's relevance, burden, and timeliness objections. The reasoning is the part employers should sit with:

Workday uses the same AI tools as its customers, and thus, under either their agent or direct-employer theory, Workday's EEO-1 and OFCCP documents are relevant to its knowledge of potential demographic disparities when utilizing AI tools.

The employer's own workforce record is in scope when AI tools are in use. The same logic applies symmetrically: if you are using an AI screening tool, your own outcome data is the discoverable record.

Why Your Tool Choice Is the Risk You Control

It is important to highlight that this is one judge's discovery order in one case. A different judge, a different vendor, a different contract, or different facts could land differently. But the pattern this ruling sets (and the uncertainty it introduces) is itself the operative point:

The vendor's bias-testing data, in Workday's case, was held privileged. If a court in your case reaches a similar conclusion about your vendor, you cannot subpoena that data to argue "the vendor tested for bias and the tool was clean." That defense may be closed to you.
The same privilege that blocks plaintiffs blocks you. If your vendor's bias testing is privileged, you cannot put it into evidence as part of your own defense.
Your applicant data, your hiring outcomes, your EEO-1 filings — all discoverable from you regardless of how the vendor's privilege fight goes. If an AI tool sits in your funnel, your workforce record will be read for whether that tool produced disparate impact.

The asymmetry is the point. The vendor's black box might stay a black box, in both directions. Your funnel data will not. And critically, you cannot predict in advance which way a future court will rule on the vendor's privilege claim, which means you cannot rely on the vendor's bias testing being available to defend you when it counts.

That points to a single conclusion: your choice of tool, and how much decision authority you give it, is the part of your risk you actually control. Tools that automate the screening decision concentrate the risk on you, because the agent-theory argument gets stronger and you have no offsetting record you can reliably point to. Tools that keep a human reviewer in the deciding seat, namely where the AI is input, not decision, break that argument and put your hiring process back in your own hands.

What This Means for Tool Selection

If you are choosing or reviewing an AI screening tool today, four questions to ask are...

1. Does the tool make the screening decision, or inform a human who does?

If the tool auto-rejects below a score threshold or auto-advances above one, your recruiter is acting on the tool's decision, not making their own. The agent-theory argument lands easily, and the legal record is the tool's output applied to your applicant pool. If a human reviewer sees the underlying material (e.g., the resume, the response, the work sample) and forms an independent judgment, the tool is input and the recruiter is the decision-maker.

That is the architecture you want.

2. What does the human reviewer actually have access to?

A reviewer who only sees a score cannot meaningfully override the tool. A reviewer who sees the candidate's actual material can.

Ask: when the recruiter looks at a candidate, do they see the score, or do they see the candidate?

3. How is the vendor's bias testing structured?

Workday's attorney-curated testing of Candidate Skills Match was held privileged. Its regulator-facing audit of Spotlight was not.

Ask your vendor which model their bias testing follows. The answer tells you whether their testing is likely to be evidence in your case, or invisible to it.

4. Who owns the applicant data?

The Mobley order turned on the Master Subscription Agreement giving customers ownership of their data. The result in this ruling is that Workday is not on the hook to produce customer applicant data, but each customer is on the hook to produce its own.

If your contract makes you the data owner, you could be the discovery target. Treat your applicant data accordingly.

Disclaimer

**This article is general commentary and does not constitute legal advice. Employers facing specific compliance questions should consult qualified counsel.**

Works Cited

Mobley v. Workday, Inc., Order on Discovery Disputes, Case No. 23-cv-00770-RFL (LB) (N.D. Cal. May 29, 2026). https://blogs.duanemorris.com/classactiondefense/wp-content/uploads/sites/56/2026/06/2026.05.29-Privilege-Bias-Test-Ruling.pdf

Mobley v. Workday, Inc., docket and pleadings (N.D. Cal. 3:23-cv-00770). https://www.courtlistener.com/docket/66831340/mobley-v-workday-inc/

Duane Morris LLP. "California Federal Court Clarifies Limits On AI Bias Testing And Applicant Data Disclosure In Mobley v. Workday." *Class Action Defense Blog*, June 2, 2026. https://blogs.duanemorris.com/classactiondefense/2026/06/02/california-federal-court-clarifies-limits-on-ai-bias-testing-and-applicant-data-disclosure-in-mobley-v-workday/

Davis Wright Tremaine. "AI Screening Tools Under Scrutiny: Federal Court Preliminarily Certifies ADEA Collective Action." *DWT Employment Law Blog*, May 21, 2025. https://www.dwt.com/blogs/employment-labor-and-benefits/2025/05/ai-hiring-age-discrimination-federal-court-workday

Seyfarth Shaw LLP. "Mobley v. Workday: Court Holds AI Service Providers Could Be Directly Liable for Employment Discrimination Under 'Agent' Theory." https://www.seyfarth.com/news-insights/mobley-v-workday-court-holds-ai-service-providers-could-be-directly-liable-for-employment-discrimination-under-agent-theory.html