TL;DR: The Workday AI hiring lawsuit, Mobley v. Workday, No. 3:23-cv-00770-RFL (N.D. Cal.), is a proposed class action in which plaintiffs allege that Workday's AI screening tools discriminated on grounds including race, age and disability. The court preliminarily certified an age-discrimination collective in May 2025, and plaintiffs filed their class-certification motion on 14 September 2026. An opposition was filed on 10 November 2026, the response deadline, and the plaintiffs' reply is due 8 December 2026, with a hearing on 9 March 2027. No court has found Workday liable.
Updated 11 November 2026. Docket checked on 11 November 2026; the last filing seen was dated 10 November 2026. What changed: on 10 November 2026, the response deadline, the docket recorded an opposition/response to a motion (Dkt 449) with supporting declarations, exhibits, a request for judicial notice and sealing motions; the plaintiffs' reply is due 8 December 2026 (docket).
Key takeaways
- The court's scheduling order of 13 July 2026 set the class-certification motion for 14 September 2026; plaintiffs filed it that day. An opposition/response was filed on 10 November 2026 (Dkt 449), replies are due 8 December 2026 and the hearing is on 9 March 2027 (N.D. Cal., Judge Rita F. Lin; order on FindLaw; docket).
- On 12 July 2024 the court granted in part and denied in part Workday's motion to dismiss; it is reported as allowing the claim that Workday acted as employers' "agent" to proceed. That was a ruling on the pleadings, not a finding of liability.
- Plaintiffs have asked the court to certify classes. These are requests, not rulings.
- The opt-in deadline for the age-discrimination collective was 7 March 2026; the court's 27 March 2026 order refers to it (docket).
- The case does not decide whether any employer's use of Workday tools was unlawful. It shapes who can be named in a suit and what records get produced.
This page answers three questions: where the case stands, what the court has actually decided, and whether any of it changes what an employer should do. A scheduling order now fixes the dates that matter through March 2027, and a discovery order of 7 October 2026 shows how the court is handling records.
Mobley is one of several matters in the wider body of AI hiring litigation; the tracker lists each case with its current status. This page stays on the Workday docket, separates allegations from rulings, and ends with a map from each ruling to a control an employer can check.
What is Mobley v. Workday about?
Mobley and others v. Workday, Inc. is a federal case, No. 3:23-cv-00770-RFL, pending in the U.S. District Court for the Northern District of California before Judge Rita F. Lin (docket). The plaintiffs are job applicants. Workday, the software vendor, is the defendant. The employers who used Workday's tools are not parties.
Everything in this paragraph is an allegation. Plaintiffs allege that Workday's AI-assisted screening tools produced a disparate impact on applicants on the basis of race, age and disability. The claims are framed under Title VII of the Civil Rights Act, the Age Discrimination in Employment Act (ADEA) and the Americans with Disabilities Act (ADA). Plaintiffs filed a Fourth Amended Complaint on 10 July 2026, so earlier descriptions that treat the Third Amended Complaint as the operative pleading are out of date (docket, Dkt 390). According to Duane Morris (24 June 2026), the court on 22 June 2026 granted in part and denied in part Workday's motion to dismiss the Third Amended Complaint (Duane Morris).
Workday disputes the case: it has moved to dismiss the claims, and the docket shows an opposition/response filed on 10 November 2026, the deadline for responses to the class-certification motion.
The case matters for one reason that does not depend on how it ends: it tests whether a software vendor, and not only the employer, can be a defendant when an AI screening tool is alleged to discriminate.
Where the Workday lawsuit stands (status box and timeline)
Status box
| Item | Status |
|---|---|
| Research date | Docket read 11 November 2026 (last filing seen 10 November 2026) |
| Court and judge | N.D. Cal., Judge Rita F. Lin, No. 3:23-cv-00770-RFL |
| Age-discrimination collective | Preliminarily certified 16 May 2025 (Dkt 128; docket; Akin Gump) |
| Opt-in deadline | 7 March 2026, as referred to in the 27 March 2026 order (Dkt 284) |
| Class-certification motion | Filed 14 September 2026 (Dkt 418), as the 13 July 2026 order required |
| Responses | Due 10 November 2026; opposition/response filed that day (Dkt 449), with supporting declarations and exhibits (docket) |
| Plaintiffs' reply | Due 8 December 2026 (Dkt 418) |
| Hearing | 9 March 2027, 10:00 a.m., Courtroom 15 |
| Liability | No court has found Workday liable |
The 13 July 2026 order reset the schedule. The order records the court's stated reason: Workday produced three bias-evaluation reports near the prior deadline (FindLaw, read 15 October 2026).
Timeline
| Date | Event | Source: docket entry |
|---|---|---|
| 12 Jul 2024 | Court grants in part and denies in part Workday's motion to dismiss (Dkt 80); the agent ruling, covered below | Docket 66831340 |
| 16 May 2025 | ADEA collective preliminarily certified (Dkt 128) | Docket 66831340 |
| 6 Mar 2026 | Court grants in part and denies in part Workday's motion to dismiss and strike the Second Amended Complaint (Dkt 267) | Docket 66831340 |
| 7 Mar 2026 | Deadline for opt-in consent forms, referred to in the 27 March order (Dkt 284) | Docket 66831340 |
| 27 Mar 2026 | Third Amended Complaint filed (Dkt 287); Workday moves to dismiss it on 10 Apr 2026 (Dkt 301) | Docket 66831340 |
| 22 Jun 2026 | Court grants in part and denies in part Workday's motion to dismiss the Third Amended Complaint, as reported by Duane Morris | Duane Morris, 24 Jun 2026 |
| 10 Jul 2026 | Fourth Amended Complaint filed (Dkt 390) | Docket 66831340 |
| 13 Jul 2026 | Scheduling order, issued after Workday produced three bias-evaluation reports near the prior deadline | Order (FindLaw) |
| 14 Sep 2026 | Plaintiffs' motion to certify class filed (Dkt 418), on the deadline the scheduling order set | Docket 66831340 |
| 5 Oct 2026 | Last day to depose experts disclosed with the motion, per the scheduling order (Dkt 395) | Docket 66831340 |
| 7 Oct 2026 | Magistrate Judge Beeler issues a discovery order granting in part and denying in part (Dkt 435) | Docket 66831340 |
| 3 Nov 2026 | Magistrate Judge Beeler denies Workday's discovery letter brief for a protective order (Dkt 445) | Docket 66831340 |
| 10 Nov 2026 | Response deadline for the class-certification motion; opposition/response filed (Dkt 449), with declarations, exhibits, a request for judicial notice and sealing motions (Dkt 450–494) | Docket 66831340 |
| 8 Dec 2026 | Plaintiffs' reply due (Dkt 418) | Docket 66831340 |
| 9 Mar 2027, 10:00 a.m. | Class-certification hearing | Docket 66831340 |
One entry in that table, the July 2024 order, bears on who can be sued.
What did the court hold on the "agent" theory?
On 12 July 2024 the court granted in part and denied in part Workday's motion to dismiss, and is reported as allowing the claim that Workday acted as an agent of employers to proceed. That is the whole holding at this stage, and it is narrower than some coverage has made it sound.
Two things are separate here, and the page keeps them apart:
- Plaintiffs allege that Workday acted as an agent of the employers who used its tools.
- The court held only that the claim could go forward. It did not decide that Workday was an agent, and it did not decide that any tool discriminated.
The ruling matters to employers because it concerns who can be named in a case about a hiring tool, and therefore whose contracts and records come under examination. It is a ruling on the pleadings. A court takes the plaintiffs' allegations as true for that purpose, so the order says nothing about whether the allegations are correct.
The "agent" concept also appears in regulatory text, under a separate regime from this federal case. The same idea is covered on the page about California's FEHA agent rule.
Does the case cover HiredScore, and who is in the proposed class?
The proposed class, as plaintiffs describe it, covers applicants "screened, ranked, scored, shortlisted, or rejected" through Workday AI tools, including HiredScore Spotlight, Fetch and Candidate Skills Match (Duane Morris, 25 Sep 2026). That is a description from a law-firm summary of the class-certification filing. Whether HiredScore tools fall inside the case, and how far, is a question for the court's orders and not for either side's marketing.
Duane Morris describes plaintiffs as proposing four primary subclasses, on the information available to it: race (African American applicants), sex (women), age (40 and over) and disability. These are requests. The court has not certified any of them, and the hearing date is in the status box.
Duane Morris also reports plaintiffs' argument that the timing of rejections suggests automated decisions. That is plaintiffs' argument and nothing more at this stage.
For applicants: the opt-in deadline passed on 7 March 2026 (Dkt 284). This page does not offer enrollment or claim guidance.
Disparate impact, explained with worked numbers
Disparate impact is a theory under which a practice that looks neutral is challenged because it falls disproportionately on a protected group. According to EEOC guidance on the Uniform Guidelines, the theory does not depend on proof that anyone intended the result (EEOC). Plaintiffs in Mobley invoke it under Title VII, the ADEA and the ADA.
The usual first test is a comparison of pass rates. The four-fifths rule treats a group's selection rate below 80 percent of the highest group's rate as a signal that the step deserves a closer look (29 C.F.R. 1607.4(D)). The ratio is called the impact ratio. A vendor or employer can run the comparison on any screening step: applications received, applications advanced, by group.
Illustrative figures, not from the case. Suppose a screening step receives 400 applications from people under 40 and 300 from people 40 and over.
| Group | Applied | Advanced | Selection rate |
|---|---|---|---|
| Under 40 | 400 | 240 | 60% |
| 40 and over | 300 | 105 | 35% |
The impact ratio is 35 / 60, or about 0.58. A ratio under 0.80 would not prove discrimination. It would tell the employer the step needs review: the job-relatedness of the criteria, the data the tool used and whether a less discriminatory alternative exists.
Running this test routinely, with a documented method, is what independent bias audits formalize: a bias audit is a structured comparison of how a tool's outcomes differ across demographic groups, performed by someone independent of the vendor or employer that built or runs the tool.
Discovery and what it means for employers' records
Discovery is where questions about records come up: what applicant data plaintiffs must produce, and on what schedule. The docket shows a discovery order from Magistrate Judge Beeler on 7 October 2026 (Dkt 435), granting in part and denying in part, and earlier discovery rulings (docket). Read the orders themselves for what they cover.
What discovery in this case does not do is reach employers. Employers are not defendants, and the case does not itself require their records. Whether an employer's data could ever be reached, for example through discovery directed at a party or a subpoena to a third party, is a question for counsel and depends on the facts. This page does not say that any employer's compliance data is "in play"; the docket entries reviewed do not support that statement.
The practical lesson is narrower. A case that turns on applicant data and bias-evaluation reports is a case about records, and the quality of an employer's own records decides how easily it can answer a question about its screening process.
The ruling-to-control map and the contract clauses to review
The table below pairs each ruling or filing with an employer control that responds to it. It describes controls to consider, not legal conclusions.
| Ruling or filing | What it concerns | Control to check | What to keep |
|---|---|---|---|
| Agent ruling (12 Jul 2024) | Who can be named in a suit over a hiring tool | Contract allocation of responsibility between employer and vendor; cooperation clauses | Executed contract, order forms, amendments |
| Disparate-impact theory | Whether a screening step affects groups unevenly | Bias testing on each screening step; decision records | Pass rates by group, test method, dates |
| ADEA collective certification (16 May 2025) | Which applicants can proceed together | Feature inventory: which AI features are enabled, since when | Configuration history, enablement dates |
| Discovery orders (for example 7 Oct 2026) | What data gets produced and how | Records retention and export | Retention schedule, export test results |
| Scheduling order (13 Jul 2026) | Bias-evaluation reports produced near the deadline | Ask the vendor which bias evaluations exist and when each was run | Vendor reports with run dates |
Contract clauses to review. Review whether the contract:
- identifies which AI features are enabled for your account and lets you switch them off;
- obliges the vendor to cooperate if a court or regulator asks for records about your use of the tool;
- allocates responsibility between you and the vendor for claims arising from automated screening;
- requires the vendor to share bias-testing results and the date each test was run;
- lets you export decision records and candidate-level data in a usable format;
- sets retention periods for decision records and says what happens to them at termination;
- requires notice when the vendor changes how a scoring or ranking feature works.
Questions for any vendor. Does the tool score or rank candidates, or only gather and present information? Who decides whether a candidate advances or is rejected, and is that configurable? What data inputs does it use? What does it keep for each decision? The answers decide which of the controls above matter most. For the wider evaluation criteria, see the guide to AI interview software used for screening.
How Mobley sits beside other cases and statutes
Detail on each of these lives on the tracker. In brief:
- Harper v. Sirius XM Radio (E.D. Mich., No. 2:25-cv-12403): on 30 September 2026 the court granted Sirius XM's motion for judgment, and on 27 October 2026 the plaintiff filed a notice of appeal to the 6th Circuit (docket). The decision is therefore being appealed.
- Kistler v. Eightfold AI (N.D. Cal., No. 3:26-cv-01768, Judge Yvonne Gonzalez Rogers) is pending. Eightfold filed a motion to dismiss on 20 April 2026, the court vacated the hearing on 24 July 2026, and the docket showed no ruling as of 13 October 2026 (docket). It concerns a consumer-reporting statute (the Fair Credit Reporting Act) rather than bias, and the Eightfold FCRA case has its own page.
- Statutory rules in New York City, Illinois, California and Colorado address notice, audits and records alongside these suits; the tracker links to them.
- For documented bias in hiring tools beyond litigation, see the collected AI hiring bias examples.
A 90-day review before the March 2027 hearing
A review started now finishes before the hearing. This is a planning outline, not legal advice; confirm each step with counsel.
Days 1 to 30: inventory which AI features are enabled
List every AI feature switched on in your applicant tracking and hiring workflow, when it was enabled, and who configured it. Note which of those features score, rank, shortlist or reject, and which only gather information for a person to review. Record which roles use each feature.
Days 31 to 60: run your own numbers
Pull pass rates by group for each screening step you can measure, using the method in the worked example above. Document the data source, the date range and who ran the comparison. Where a step shows a ratio under 0.80, note what you checked next and what you decided.
Days 61 to 90: review contracts and train reviewers
Walk the contract clause list above against each vendor agreement. Then train the people who review AI output so they know what the tool shows, what it does not show, and how their own decision is recorded.
Should employers stop using Workday AI screening? That is a question for counsel, not a recommendation from this page. The case does not prohibit use. What it points to is configuration, records and contract terms.
FAQ
What is the current status of Mobley v. Workday?
Class certification is being briefed: plaintiffs filed their motion on 14 September 2026, an opposition was filed on 10 November 2026, the plaintiffs' reply is due 8 December 2026, and the hearing date is in the status box. An age-discrimination collective was preliminarily certified earlier. No court has found Workday liable.
Is Workday liable, or is the employer?
No court has found either liable. The July 2024 agent ruling, covered above, was on the pleadings and is not a finding. How any employer is affected is a question for counsel.
Does the lawsuit cover HiredScore?
Plaintiffs describe the proposed class as including applicants screened through HiredScore tools, among other Workday AI tools, according to Duane Morris. How far the case reaches is a question for the court.
Is this a class action? Can applicants still join?
An age-discrimination collective was preliminarily certified. The opt-in deadline for the collective was 7 March 2026. A class-certification motion on further proposed classes, filed 14 September 2026, is pending. This page gives no claim or enrollment guidance.
Should we stop using Workday AI screening?
The case does not prohibit use. What matters is how the tool is configured, what records exist and what the contract says. Ask counsel before changing a screening process.
What is disparate impact in AI hiring?
Disparate impact is a challenge to a neutral-looking practice that falls disproportionately on a protected group. In hiring, it is tested by comparing selection rates between groups. The page on independent bias audits covers the method.
Can our company's data be pulled into the case even though we are not a defendant?
Employers are not defendants, and the case does not itself require their records. Whether an employer's data can be reached, for example through discovery directed at a party or a subpoena to a third party, depends on the facts and is a question for counsel.
What should we ask our AI hiring vendor now?
Ask which AI features are enabled for your account, what bias-testing results exist and when each was run, how you can export decision records, and what cooperation the contract requires if a court or regulator asks for records.
How does this relate to NYC Local Law 144 and the EU AI Act?
Both are statutory regimes that sit apart from the discrimination claims in this case. The tracker and the compliance pages carry the detail on each.
Run the records test on your own process
The orders in this case turn on records. See what a decision record with reasoning and a transcript looks like: See an evidence pack in a demo.
Comments
No comments yet.
Comments are closed on this post.