AI Resume Screening Lawsuit: Mobley v. Workday

Short answer: the case everyone means by "the AI resume screening lawsuit" is Mobley v. Workday, Inc., No. 3:23-cv-00770-RFL, in the U.S. District Court for the Northern District of California. It matters to employers for one reason: the court let discrimination claims proceed against the software vendor on the theory that a screening vendor can act as an agent of the employers using it. In May 2025 the court preliminarily certified a nationwide age-discrimination collective, and court-authorized notice went out with an opt-in deadline of March 7, 2026. As of July 2026 the case is in discovery, two significant rulings landed in June 2026, and no trial date has been set.

Last updated July 2026. This is a plain-language summary for hiring teams, not legal advice. Talk to your employment counsel about your own tools and process.

What is the Workday AI lawsuit about?

Derek Mobley sued Workday alleging that its applicant-screening tools disproportionately rejected applicants who were Black, over 40, or disabled. He described applying to a large number of roles through employers running Workday and being rejected repeatedly, often within minutes of applying and outside of business hours, which is the pattern that suggests a machine and not a person made the call.

The unusual part is who got sued. Mobley did not sue the hundreds of employers whose job postings he applied to. He sued the software company. Workday's defense was the obvious one: we are a vendor, we do not employ anyone, and federal employment statutes apply to employers. The court did not accept that framing. It held the complaint adequately alleged Workday acts as an agent of its client employers when it performs a function the employer would otherwise perform itself, which brings it inside the coverage of Title VII, the ADEA, and the ADA.

A later ruling in July 2025 expanded the scope to cover the HiredScore AI products as well, so this is not narrowly about one screening feature. HiredScore is the candidate grading engine Workday acquired in March 2024 and now sells as HiredScore AI for Recruiting; if you are trying to work out whether your own contract includes it, our breakdown of Workday AI resume screening walks through what each capability does and how to check. Teams weighing whether to keep that screening inside the platform or run it separately usually end up comparing a Workday alternative for the screening step alone.

Who is included in the Workday lawsuit?

The collective the court preliminarily certified on May 16, 2025 covers all individuals aged 40 and over who, from September 24, 2020 through the present, applied for jobs through Workday's application platform and were denied employment recommendations. The court approved the notice plan on December 2, 2025 and set an opt-in deadline of March 7, 2026, which has now passed, so membership is settled.

The scale is what made this national news. Workday's own filings put roughly 1.1 billion applications rejected through its system across the covered period, and the eligible collective has been described as running to millions of applicants. The number who actually opted in by the deadline is far smaller: around 14,000 people. That gap is worth noticing, because the headline figure and the real collective are two very different things, and most coverage conflates them.

Only the ADEA age claim is certified as a collective. The race and disability claims remain in the case but have not been certified as a class.

Can an AI vendor be sued for discrimination?

On the current state of this case, yes, at least far enough to reach discovery. That is the holding that changed how careful buyers read their vendor contracts. The theory is not that the software company is your employer; it is that when a vendor takes over a decision the employer would otherwise make, the vendor is acting as the employer's agent and cannot hide behind being "just a tool."

The practical consequence cuts both ways, and the second direction is the one hiring teams underweight. If the vendor can be an agent of the employer, the employer is on the other end of that relationship. Nothing in this case makes an employer less liable for a screening tool it chose, configured, and relied on. "The model did it" was never a defense under Title VII, and this litigation has made that concrete enough that general counsel are now asking about it.

Is AI resume screening legal?

Yes. Using AI to screen resumes is legal in the United States, and nothing in Mobley says otherwise. What is illegal is discrimination, whether a human or a model produces it. Disparate impact liability under Title VII, the ADEA, and the ADA applies to an algorithm exactly as it applies to a hiring manager with a bad habit, and an employer does not need to intend a discriminatory outcome to be liable for one.

One point of confusion worth clearing up: the EEOC withdrew its May 2023 technical assistance document on AI and Title VII in January 2025 following Executive Order 14179. Some vendors have quietly read that as deregulation. It is not. Withdrawing guidance does not repeal a statute. The underlying law and disparate impact liability are untouched, and this lawsuit is being litigated under those statutes right now. Meanwhile state and city rules have gone the other way and multiplied, which we track in our rundown of AI hiring laws by state and in detail for NYC Local Law 144 bias audits.

What does this mean for employers using AI hiring tools?

It means the question your counsel will ask is no longer "do we use AI in hiring" but "what exactly does it decide, on what criteria, and can you show the reasoning for any individual candidate." Most teams cannot answer the third part, and that is the exposure.

Three specifics are worth checking this quarter, because they are the ones that separate a defensible process from an indefensible one.

Does the tool reject anyone without a human? This is the single biggest fork. A tool that ranks candidates and flags why is an aid to a human decision. A tool that auto-rejects on a knockout answer is making the employment decision, and a badly worded question then removes qualified people with nobody watching. If any tool in your stack auto-rejects, know it, and be able to justify every knockout as job-related.

Are the criteria job-related and written down? Criteria that are documented, applied identically to every applicant, and tied to what the role actually requires are defensible. Vague "culture fit" scoring and proxies for age (graduation year, "digital native", years-since-degree bands) are not. We wrote a practical guide to drafting these in resume screening criteria, including the wording tests that catch a proxy before it ships.

Can you reconstruct a specific decision? A year from now, if a rejected applicant asks why, can you produce the criteria that applied, the evidence in their application, and the reason they ranked where they did? If the answer is a score with no explanation, you have bought a black box, and the discovery phase of this case is a preview of what happens when someone asks a court to open one. Keeping a live register of which automated tools touch employment decisions, what each one does, and which state rules attach to it is now ordinary hygiene, and teams that already track their obligations and controls in one place tend to find this exercise takes an afternoon rather than a quarter.

How do I reduce legal risk when using AI resume screening?

Keep the human in the loop, document the criteria, disclose the AI use to candidates, get a bias audit, and keep records. In practice that is five concrete controls, and none of them require you to stop using screening software.

The reason to do this now rather than after a complaint is that four of the five are cheap while your process is small and expensive once you have two years of undocumented decisions behind you. Our longer walkthrough on how to reduce bias in AI resume screening covers the audit mechanics, and EEOC standards and AI hiring fairness covers what the statutes actually require now that the 2023 guidance is gone.

Does this mean we should stop using resume screening software?

No, and the argument that it should is weaker than it first looks. Manual screening is not a neutral baseline. A recruiter working through 400 applications gives the first 50 real attention and skims the rest, applies a different standard on a Friday afternoon than a Monday morning, and leaves no record of why anyone was cut. That process is not more defensible than an automated one. It is just harder to audit, which is not the same thing as being fair.

What the litigation actually rewards is a specific shape of tool: one that applies consistent, job-related criteria, writes down its reasoning, discloses itself to candidates, and stops short of making the decision. That shape is more defensible than either a black-box scorer or an unrecorded human skim. If you are evaluating vendors on this basis, our comparison of the best resume screening software notes which tools auto-reject and which publish their reasoning, and the AI resume screening software page explains how the ranking and reasoning work on a live applicant pile.

What to do this week

Pull the list of every tool that touches an application before a human sees it, including the filters and knockout questions inside your ATS, which teams routinely forget are automated screening. For each one, write down what it decides, what criteria it uses, and whether it can reject candidates on its own. Then fix the ones that decide without a person, and document the ones that do not.

That inventory is the thing counsel will ask for first, and almost nobody has it. Building it takes an afternoon and it is the same document that makes a bias audit, a Local Law 144 filing, and a vendor security review straightforward instead of frantic. If your screening criteria are the part that is undocumented, the applicant eligibility screening page shows what a written, per-role eligibility gate looks like when it is set up properly.