DEV Community

Multigrid
Multigrid

Posted on • Originally published at multigrid.ai

The Workday AI Hiring Lawsuit: Where the Case Stands

Every ruling in Mobley v. Workday so far has been about whether the case can proceed and against whom. None has been about whether a Workday tool discriminated against anyone. Keeping those apart is the whole point of reading the docket rather than the headlines.

The case, and its posture

Mobley v. Workday, Inc. is pending in the United States District Court for the Northern District of California, case number 3:23-cv-00770, before Judge Rita F. Lin. Derek Mobley filed in February 2023, alleging that he had applied for many roles through employers using Workday’s applicant screening products and been rejected each time, and that the screening discriminated on the basis of race, age and disability. The claims are brought under Title VII of the Civil Rights Act, the Age Discrimination in Employment Act, the Americans with Disabilities Act and 42 U.S.C. § 1981. The public docket is on CourtListener, and it is the place to check any statement on this page against the filings.

What makes the case unusual is the defendant. Mobley did not sue the employers who rejected him. He sued the software vendor whose product they used, which meant the threshold question was not whether discrimination occurred but whether the federal employment statutes reach a company that is nobody’s employer.

A posture summary of a live case as at the date on this page, not legal advice and not a prediction. Rulings in active litigation are superseded quickly — verify against the docket, and take advice on your own exposure rather than inferring it from another party’s procedural history.

The agent ruling, and its limits

The court dismissed the first complaint in January 2024 with leave to amend. On 12 July 2024 it ruled on the amended complaint, and that order is the one everything else rests on.

The court rejected the theories that Workday was itself an employer or an employment agency, and accepted a third: that Workday could be liable as an agent of its employer clients. The reasoning is that the federal anti-discrimination statutes define an employer to include agents, and that where an employer delegates a traditional hiring function — screening, ranking, rejecting — to a third party, that third party can be performing the employer’s function and can be answerable for it. The allegation that Workday’s tools did not merely pass applications along but recommended rejections was central.

The limits of that holding are strict and routinely overstated. It is a ruling on a motion to dismiss, so the court assumed the allegations were true and decided only that they state a claim if proved. It is a district court decision, binding on nobody, and no appellate court has endorsed it. It says a vendor can be an agent on facts like these; it does not say Workday is one, and it makes no finding about any tool’s behaviour. What it does do is remove the argument that software vendors sit outside employment law entirely — which was, until 2024, a serious argument.

The ADEA collective and its scale

On 16 May 2025 the court granted preliminary certification of a nationwide collective on the ADEA claim, covering applicants aged 40 and over who applied through employers using the relevant Workday features and were rejected. The ADEA uses the Fair Labor Standards Act’s opt-in collective mechanism rather than Rule 23 opt-out classes, which means membership depends on people affirmatively joining, and the practical size of the case is decided by notice and response rather than by the certification order.

Scale is what makes the case matter commercially. Filings in 2025 indicated that the number of applications rejected through the relevant products in the covered period ran to roughly 1.1 billion, and Workday raised the prospect of a collective numbering in the hundreds of millions of people. Those are figures about applications and potential members, not about anyone found to have been discriminated against, and the distinction is easy to lose in reporting.

In July 2025 the court ordered Workday to produce a list of the employers that had enabled the HiredScore AI features, and the collective definition was framed to reach applicants whose applications were scored, sorted, ranked or screened using them. Notice to potential members was authorised in February 2026 with an opt-in deadline of 7 March 2026, so the membership of the collective is now closed and known to the parties.

The 2026 rulings

Three developments, in order.

  • January 2026. The court granted leave to amend to add further named plaintiffs and additional claims, including sex-based claims under Title VII and claims under the California Fair Employment and Housing Act.
  • 6 March 2026. The court rejected Workday’s argument that the ADEA’s disparate impact provision does not protect job applicants, holding that applicants aged 40 and over may bring disparate impact claims. This closed off what had become Workday’s strongest remaining threshold defence on the age claims. It is a ruling on the scope of a statute, not on the tools.
  • 22 June 2026. The court denied in part a motion to dismiss the fourth amended complaint, leaving claims across race, sex, age and disability in the case.

The case remains in discovery. That is the stage where the substantive questions — what the tools actually do, what testing exists, what the outcome data shows — get answered, and none of those answers is public.

What no court has decided

The most useful paragraph on this page is the list of things that remain entirely open after three and a half years of litigation.

Whether any Workday tool discriminated against anyone.There has been no trial, no summary judgment on the merits, and no factual finding about the products. Every ruling described above is procedural.

Whether the agent theory survives contact with the evidence. Surviving a motion to dismiss means the allegations state a claim. Whether the relationship in fact amounts to delegation of a hiring function is a question for summary judgment or trial, on contracts and configurations nobody outside the case has seen.

Whether the theory would survive appeal. A district court order is not precedent. The Ninth Circuit has not ruled, and other circuits are not bound by anything here.

How liability would divide between vendor and employer.The employers are not defendants. If a screening tool produced an unlawful impact, the allocation between the company that built it and the company that switched it on is untouched by anything decided so far — which is precisely the risk that allocation clauses in AI vendor contracts are trying to price.

What would actually change the picture

Four events, and it is worth knowing which reporting to take seriously.

A merits ruling — summary judgment or trial — would be the first decision about the technology rather than about procedure. An appellate decision on the agent theory would convert a district court’s reasoning into law in at least one circuit. A settlement would end the case without deciding anything, which given the scale is a realistic outcome and would leave the July 2024 order as the most cited thing to come out of it. And a statutory intervention, federal or state, could reallocate vendor liability directly; several state legislatures have taken up the question in different forms, surveyed in the state AI law landscape.

In the meantime, the defensible reading is narrow. A federal court has held that a screening vendor can be sued as an agent of the employers that use its product, and a large age-based collective is proceeding on that basis. Nothing has been found. Anyone telling you the case establishes that AI hiring tools are unlawful, or that vendors are liable, is describing a case that has not happened yet — and the same caution applies to the contrary claim that the EEOC’s iTutorGroup settlement shows the exposure is limited to explicit rules.

Related

Top comments (0)