The Workday Class Action Lawsuit: Is Your AI at Risk?

For years, the legal burden for hiring discrimination fell squarely on the employer. The Workday class action lawsuit could change that permanently. The case argues that the AI isn't a passive tool following instructions — it's an active participant in the decision, screening, ranking and scoring candidates before a human ever sees them. That raises the question the whole industry now has to answer: if an algorithm contributes to a biased outcome, can the vendor be held liable? For any organization that builds, sells or uses AI in hiring, this case is a wake-up call. It is no longer enough to buy a tool and assume the end-user carries all the risk.

Key Takeaways

• Shared responsibility is the new standard. The lawsuit shows that AI vendors and employers can both be held accountable for biased hiring outcomes — the legal risk now extends beyond the company making the final hire.
• You must be able to prove your AI is fair. Trusting your tools is not a defense. You need objective evidence from regular bias audits and clear documentation to stand behind your hiring process under legal and regulatory scrutiny.
• A proactive governance plan is essential. A defensible AI strategy takes more than technology — it demands meaningful human oversight, rigorous vendor vetting, and clear internal policies that build fairness in from the start.

What the Workday Lawsuit Is About

If you work in HR or buy HR technology, you have heard of the class action against Workday. The case, Mobley v. Workday, Inc. (N.D. Cal., Case No. 3:23-cv-00770), strikes at the heart of how modern hiring works: it questions whether the AI tools meant to make recruitment more efficient are quietly building unfair barriers for qualified candidates. It is not a problem for one company — it is a test for the entire ecosystem of AI-powered hiring platforms, and it puts vendors and employers on notice together.

The core allegation: AI-driven hiring bias

The central claim is that Workday's AI screening tools create a disparate impact — an outcome where applicants from protected groups are filtered out at a higher rate, even without any explicit instruction to discriminate. The lawsuit alleges the tools disproportionately screen out applicants based on age (over 40), race (specifically African American candidates), and disability. The argument is that the AI learned and replicated historical patterns of discrimination, effectively denying qualified people the right to compete on equal footing.

The Workday tools named in the case

The suit doesn't make a vague claim against “AI.” It names specific features inside Workday's platform. The Candidate Skills Match system automatically compares an applicant's qualifications against a job posting to save recruiters time; the lawsuit alleges it unfairly filters out older candidates, applicants with disabilities, and African American applicants. The Workday Assessment Connector, which integrates third-party skills tests and personality assessments into the platform, is also under scrutiny for how it processes and weights those results. Knowing which tools are named is the first step to assessing the same exposure in your own stack.

The age discrimination (ADEA) claim

The age claim is the part of the case that has advanced furthest, and it is instructive because it shows how an algorithm can discriminate by age without ever being told an applicant's age. AI hiring tools are pattern-matchers: if a company has historically hired younger workers, the model learns to associate youth-adjacent signals — graduating within the last ten years, proficiency in newer software, the absence of career gaps — with being a “good” candidate. Those signals act as proxies for age. A federal court found enough reason to believe Workday's tools may have been designed using “biased training information” that reflects past hiring patterns. The lead plaintiff, Derek Mobley — who is over 40, African American, and has a disability — alleges he was rejected for dozens of jobs at companies using Workday's tools despite being qualified.

Race and disability claims

Beyond age, the lawsuit raises concerns about bias against applicants based on race and disability. Because the lead plaintiff falls into all three protected categories, the case tests whether a single automated system can systematically disadvantage Black applicants and people with disabilities by screening them out before a human review ever happens.

Who the Lawsuit Covers — and Where It Stands Now

The court has defined a specific group of people who can take part in the collective action, turning abstract concerns about AI bias into a concrete legal matter. For HR leaders, recruiters and vendors, the eligibility rules and the case timeline are the clearest look at exactly what is being put on trial.

How the Case Got Here: A Timeline

The case has been building for more than three years. The key milestones, in order:

Additional timeline details can be found in the case's full court docket and filings, CourtListener (Free Law Project).

Who was eligible to join

Eligibility centered on two conditions: being 40 years of age or older, and having applied for a job through Workday's platform between September 24, 2020 and the present. This aligns with the core allegation that the tools systematically screen out older applicants under the Age Discrimination in Employment Act (ADEA). According to court documents, Workday's software was used to process roughly 1.1 billion job applications during the covered period — which is why a single algorithm can translate into a collective action of enormous scale. A recurring piece of the plaintiffs' evidence is that many rejections arrived within minutes or hours of applying, implying an automated decision rather than a human one.

The opt-in window has closed

Eligible applicants joined by filing an “Opt-In Consent to Join Form” through the official Workday case website. That deadline — March 7, 2026 — has now passed, so the collective membership is set. If you are an employer or vendor reading this after the deadline, the takeaway is not the form; it is that the case is now moving into its substantive phase with a defined nationwide group behind it.

Where the case stands now

With the collective membership now set, the case has moved into its merits phase: discovery into how the tools score and screen applicants, and a likely fight over whether the collective stays certified. Workday may move to decertify later, but for now it proceeds as a nationwide collective with the question of vendor liability squarely in play — and the April 2026 revival of California state-law and disability claims widens the exposure beyond age alone. Early discovery rulings are already shaping the evidence: in May 2026 the court shielded Workday's attorney-curated bias-testing data under privilege, but ordered it to hand over its own EEO-1 and OFCCP records. The most consequential signal remains the court's treatment of the AI as a single, overarching policy that screens, sorts and scores every applicant it touches. This section is updated as the case develops. For background, see coverage from SHRM and the plaintiffs' counsel notice.

How Bias Creeps Into AI Hiring Tools

You adopt an AI hiring tool to be more efficient and more objective. The risk is that it quietly does the opposite. AI learns from the data it is given, and if that data reflects past human bias, the model treats those biases as the target to reproduce. This isn't a simple bug — it is a structural risk that shows up in three common ways.

Algorithms can discriminate without naming a protected trait

An algorithm follows rules to reach a decision, and those rules can produce discriminatory results without ever referencing age, race or disability. A model never told to screen out older applicants can still learn to penalize early graduation dates or long careers — reaching the same biased outcome by proxy.

Biased training data is the root cause

“Garbage in, garbage out” is especially true for AI. Feed a model a decade of hiring records that subconsciously favored certain candidates, and it will encode those preferences as correct. In the Workday case, the court found reason to believe the tools were designed “in a way that shows employer biases and uses biased training information.” Regular AI bias auditing is how you surface these patterns before they cause legal and reputational damage.

Resume screeners are especially vulnerable

Modern screeners do more than match keywords — they actively participate in the decision. That is where bias compounds: a tool can learn to associate certain names with a gender, or to penalize employment gaps tied to disability or caregiving. Gaining visibility into these systems through an AI assurance platform is essential to making fair, defensible talent decisions.

What “Disparate Impact” Actually Means

Much of this case turns on disparate impact — a legal theory where a neutral-looking practice has a disproportionately negative effect on a protected group, regardless of intent. You don't have to mean to discriminate to be liable; what matters is the outcome. For AI hiring, that is the whole ballgame: “we didn't program it to consider age” is not a defense if the tool nevertheless screens older applicants out at a higher rate. It shifts the question from “what did the algorithm intend?” to “what did the algorithm do?”

Why This Case Matters for the Future of AI in Hiring

Legal risk has reached the technology vendor

For a long time the legal burden sat with the employer. By allowing the case to proceed against Workday itself, the court has signaled that the creators of these tools may share responsibility for their impact. For HR technology vendors, “we just provide the software” is no longer a sufficient answer.

Regulators are watching the same question

The lawsuit doesn't exist in a vacuum. The EEOC has been explicit that employers are responsible for their hiring tools whether built in-house or bought from a vendor, and is focused on AI's role in employment decisions. Waiting for a complaint is not a strategy.

It is setting precedent for AI accountability

This is an early test of how foundational civil-rights law applies to modern technology. By treating an algorithm's systematic impact as seriously as a discriminatory human policy, the case lays groundwork that will shape litigation and regulation for years.

Mobley Isn't Alone: A Widening Wave of AI Hiring Litigation

Mobley is the most advanced of these cases, but it is not the only one — and reading it in isolation understates the exposure. A growing cluster of lawsuits is testing AI hiring from different legal directions, hitting both the vendors that build these tools and the employers that deploy them. Here is how they compare at a glance:

Kistler v. Eightfold AI: a new privacy and FCRA front

In January 2026, applicants Erin Kistler and Sruti Bhaumik filed a class action against Eightfold AI, alleging its hiring platform operates as an unregistered consumer reporting agency under the Fair Credit Reporting Act (FCRA). Unlike Mobley, this is not a bias claim. It targets how the AI compiles and uses applicant data — social media profiles, location, online activity — and scores or labels candidates (for example, “team player” or “introvert”) before any human review, allegedly without the disclosures, consent, and dispute mechanism the FCRA requires. The case (removed to federal court and in the pleading stage as of mid-2026) opens a second legal front: even where no discrimination is alleged, an AI hiring vendor can face liability over data collection and transparency. See the Akin AI Law tracker and Fox Rothschild's analysis.

Swanson v. IBM: the age theory aimed at the employer

Filed in May 2026 in the U.S. District Court for the Western District of Texas, Swanson v. IBM extends the age-discrimination theory — but points it at the employer's own use of AI rather than the vendor's. Daniel Swanson, a 24-year IBM veteran, alleges he was pushed out in a 2024 “Resource Action,” then rejected for a comparable role via an AI-generated rejection, as part of a strategy favoring younger “Early Professional Hires.” The complaint leans on the same ADEA logic as Mobley and cites a prior EEOC reasonable-cause determination. It shows the Mobley age-bias argument migrating from the platform maker to the companies deploying these tools. Background via Human Resources Director.

Harper v. SiriusXM: race discrimination aimed at the employer

In a suit filed in 2025, Arshon Harper alleges he applied to roughly 150 roles at SiriusXM and was rejected from all but one, and that the company's AI screening leaned on proxies for race — such as education and address — to filter him out. As in the IBM case, the defendant is the employer, not the vendor, but here the claims run under Title VII of the Civil Rights Act and Section 1981, alleging both disparate treatment and disparate impact. It is the race-discrimination counterpart to the IBM age case, and another reminder that liability extends to the organizations deploying these tools, not just the companies building them. (See our full breakdown: Harper vs. SiriusXM: The Growing Legal Risk of AI in Hiring.)

What the pattern means for you

Read together, these cases widen the risk surface on several fronts at once: discriminatory outcomes by age (Mobley, IBM) and by race (Harper), and data and privacy compliance (Eightfold) — hitting both the vendors that build these tools and the employers that deploy them. Notably, none of them hinges on new AI-specific regulation; they apply long-standing civil-rights and consumer-protection law (Title VII, Section 1981, the ADEA, the ADA, the FCRA) to new technology, which makes the exposure immediate rather than hypothetical. The common thread is automated decisions made at scale, with thin human review and little documentation to explain them. Whether you are a vendor or an employer, the defense is the same set of disciplines — independent bias auditing, transparency about how a tool collects and uses candidate data, and human oversight you can actually evidence.

What This Means for You: Employers and Vendors

These cases land differently depending on which side of the tool you sit on — but the through-line is identical: you have to be able to show your work. Here is where each role should focus.

For HR and talent acquisition teams (employers)

• You own the outcomes of the tools you buy. Liability does not stop at the vendor — Harper and Swanson both target employers directly for the AI they deployed.
• Commission an independent bias audit of any automated tool you use to screen or rank candidates — and if you hire in NYC, publish it to meet Local Law 144.
• Document your decisions. Keep clear records of why candidates are advanced or rejected; vague vendor “fit scores” you cannot explain are a liability, not a feature.
• Keep meaningful human oversight. Reviewers must be trained to question the tool and empowered to override it — a rubber stamp is not oversight.
• Vet vendors with evidence, not assurances (use the checklist below).

For AI builders and HR-tech vendors

• “We just provide the software” is no longer a defense. Mobley and Kistler put vendors directly in the legal frame — under both anti-discrimination and consumer-protection law.
• Build fairness in from the start. Test for disparate impact across protected groups before you ship and continuously after, because models drift.
• Be transparent about data. The Eightfold/FCRA case shows how you collect and use candidate data is now its own liability surface, not just bias.
• Make your customers audit-ready. Provide configuration-level results and documentation they can use to meet their own Local Law 144 and EU AI Act obligations.
• Get independently certified. A third-party standard such as Warden Assured lets buyers verify your claims instead of taking them on trust.

How to Audit Your AI Hiring Tools for Bias

The lawsuit is a clear signal that deploying a tool is not enough — you have to prove it works fairly. A thorough audit shows you what is happening inside the “black box,” surfaces risk, and lets you act before a problem becomes a filing.

Run continuous bias testing and monitoring

Treat your AI like any critical system: it needs regular check-ups. Models drift, so a one-time review is not enough. Periodically test your models against diverse datasets to check for statistical disparities across demographic groups, and monitor on an ongoing basis.

Keep compliant, legal-grade documentation

If you can't explain how your AI reached a decision, you can't defend it. Keep clear records of why candidates are advanced or rejected, the model's purpose, the data it was trained on, and the factors it weighs. Vague “fit scores” with no transparent reasoning are a liability. Warden AI's assurance platform helps generate this evidence automatically.

Bring in an independent, third-party auditor

An internal review is a start, but an independent audit gives regulators and customers objective assurance. Achieving a certification such as Warden Assured demonstrates that your AI meets a recognized standard for fairness, transparency and accountability.

How to Proactively Prevent AI Discrimination

Build in meaningful human oversight

A person in the loop only helps if that person is trained and empowered to question the tool. Treat AI as a co-pilot, not the pilot: recruiters should understand its limits, review its recommendations — especially rejections — and have the authority to override them.

Do real vendor due diligence: the questions to ask

When you use a third party's AI, their risk becomes your risk — so vet with evidence, not assurances. Ask every vendor these questions, and ask for documents, not just answers:

Bias and fairness testing

• Has an independent bias audit been completed in the last year for the exact configuration we will use — and can you share the auditor's name and a summary report?
• Which protected-class proxies did you test for, and what were the adverse-impact (selection-rate) ratios? Do they clear the EEOC four-fifths rule (an impact ratio of about 0.85 or higher)?
• How often do you re-test for bias, and what triggers a retrain?

Data and privacy

• What data were the models trained on, and what data do you collect on applicants — including any third-party or web-sourced data?
• How do you obtain consent and handle candidate data rights, and have you assessed whether your use of applicant data implicates the FCRA?

Transparency and explainability

• How are candidate scores actually calculated, and can a recruiter see why a candidate was ranked or screened out?
• Will you provide documentation we can use to defend a hiring decision — model purpose, inputs, and known limitations?

Human oversight

• What human checks are built into the workflow, and can we configure or override the tool's recommendations?

Compliance and certification

• Can you support our NYC Local Law 144 bias-audit and notice obligations?
• Have you completed — or can you support — an EU AI Act conformity assessment for high-risk employment systems?
• Do you hold an independent, third-party certification (such as Warden Assured) that we can verify?

Scope, supply chain, and operational risk

The following five questions are drawn from the CareerXroads (CXR) community's collaborative AI vendor evaluation guide — a deeper checklist worth using in a full RFP. Get the full CXR list of questions to ask AI vendors.

• Where are AI, ML, or LLM models used in your product, and what specific tasks do they perform (for example ranking, matching, or writing)?
• Do you use third-party AI providers or subprocessors — and can you provide due-diligence documentation for them?
• What incident history — breaches, misuse, or litigation — are you willing to disclose?
• How do you manage model updates, and how do you communicate model changes to us?
• What are your data deletion timelines and opt-out processes, and who owns the data or IP created from our use of the product?

Stand up an AI governance framework

A governance framework is your rulebook for using AI responsibly: clear policies, roles and procedures across procurement, deployment, monitoring and updates, plus continuous AI bias auditing to catch issues before they become systemic. For larger teams, building this into enterprise process is the difference between reacting to risk and managing it.

Keeping Your AI Hiring Compliant

Compliance comes down to a process you can stand behind. In the U.S., New York City's Local Law 144 requires employers using automated employment decision tools to commission independent bias audits and publish the results — and the responsibility sits with the employer, not just the vendor. The EU AI Act classifies hiring AI as “high-risk,” with strict transparency, risk-management and human-oversight obligations. Comparable rules are emerging at the U.S. state level.

One caveat worth tracking: a December 2025 federal executive order seeks to preempt and challenge state AI laws, so the patchwork of state requirements may shift. The underlying anti-discrimination exposure in cases like Mobley, however, is grounded in long-standing civil-rights statutes (Title VII, the ADEA, the ADA) that are not affected by that order.

Sources

• Mobley v. Workday, Inc. (3:23-cv-00770) — full court docket and filings, CourtListener (Free Law Project)
• Mobley v. Workday, Inc. — case summary, Civil Rights Litigation Clearinghouse
• Federal court authorizes notice; opt-in by March 7, 2026 — National Law Review
• Workday case update — Wiggins Childs (plaintiffs' counsel)
• EEOC — Age discrimination (ADEA) overview
• May 29, 2026 discovery order on AI bias-testing data and applicant data — Duane Morris, Class Action Defense
• Kistler et al. v. Eightfold AI Inc. (FCRA class action) — Akin AI Law tracker
• Swanson v. IBM — Human Resources Director

Related Articles

• Harper vs. SiriusXM: The Growing Legal Risk of AI in Hiring
• Case Briefing: Swanson v. International Business Machines Corporation (IBM)
• The Eightfold Class Action: 5 Webinar Insights on FCRA & AI Hiring Risk
• Evidence, Not Assumptions: Insights from our Mobley v. Workday Panel
• Age Bias in AI Hiring: Addressing Age Discrimination for Fairer Recruitment
• State of AI Bias in Talent Acquisition