AI Hiring Lawsuits 2026: Workday Class Bid & Eightfold FCRA Case

Two federal cases moved forward in the months leading into this September, and between them they sketch the shape of AI-hiring liability for the next few years. Neither is about whether an algorithm can be biased — courts have been circling that question since 2023. Both are about something narrower and, for any company that has bought an AI screening or scoring tool, more immediately useful: which body of law actually reaches the vendor, and what does a plaintiff have to show to get there.

Mobley v. Workday: from one plaintiff to "potentially millions"

Mobley v. Workday, Inc. (No. 3:23-cv-00770, N.D. Cal.) started in 2023 as a single rejected applicant's claim that Workday's AI-powered screening tools discriminated against him on the basis of race, age and disability. It has since become the case every HR-tech vendor watches, for a reason that has nothing to do with Workday specifically: it is testing whether a software vendor, rather than the employer that deployed the software, can be sued directly under federal anti-discrimination law for how its product screens candidates on behalf of thousands of separate customers.

The case has moved in two hard-edged steps. In May 2025, the court granted conditional certification under the Age Discrimination in Employment Act (ADEA) for applicants 40 and older whose applications were scored, sorted, ranked or screened by Workday's AI and did not result in a hire recommendation — the first time an AI-hiring vendor, rather than an individual employer, faced anything resembling a certified collective. On July 7, 2025, the court expanded that collective again: it held that applicants screened through Workday's HiredScore AI features were included too, and ordered Workday to identify which of its customers actually use HiredScore so those applicants could be folded into the case.

That was still an ADEA collective — age discrimination only, and procedurally a lighter lift than a Rule 23 class action. This September, the plaintiffs went further: they filed a full motion for class certification covering the broader protected characteristics at issue in the case — race, sex, age and disability — arguing the claims can be resolved class-wide because the same automated scoring pipeline is applied uniformly across customers. The scale being argued is not hypothetical: the motion frames the potential class as reaching "potentially thousands, if not millions" of job applicants who passed through Workday's screening tools over the relevant period. Part of what let the plaintiffs get there was discovery: reporting on the case notes that the certification timeline was extended by roughly two months after Workday produced its own internal bias evaluation reports on the challenged AI tools — the kind of document a vendor generates for its own risk management and then has to hand over once litigation reaches that stage. A hearing on the class certification motion is now scheduled for March 9, 2027.

Whatever the eventual ruling, the posture of the case is already instructive. The theory plaintiffs are pushing is that when one vendor's model scores applicants for hundreds of unrelated employers, the vendor — not each individual employer — is the entity whose conduct is actually common across the class, which is the whole point of a class action. If that theory survives certification, "we just bought the tool, we didn't build it" stops being a reliable place for an employer, or a vendor, to stand.

Eightfold: the same problem, a completely different statute

Kistler et al. v. Eightfold AI Inc. is a smaller case procedurally, but it reaches for a legal theory that has almost nothing to do with employment discrimination law. Two California applicants, Erin Kistler and Sruti Bhaumik, filed suit on January 20, 2026 in Contra Costa County Superior Court; Eightfold removed it to the Northern District of California on March 2, where it is now docketed as No. 3:26-cv-01768 before Judge Yvonne Gonzalez Rogers in Oakland. Bhaumik says she was screened out of a Responsible AI role at Microsoft; Kistler says she applied through a PayPal careers link that routed through Eightfold's own domain. Neither plaintiff was ever told a score existed.

The claim is not disparate impact. It's that Eightfold's "likelihood of success" score — generated on a 0-to-5 scale and used to sort applicants before a human ever sees the file — is a consumer report under the federal Fair Credit Reporting Act (FCRA) and California's parallel Investigative Consumer Reporting Agencies Act (ICRAA). Both statutes were written for background-check companies, not AI vendors, but both turn on a functional test: does a third party assemble information about a consumer's character, reputation or general qualifications, and furnish it to be used for an employment decision? If a score built from resume signals and behavioral inference meets that test, the vendor generating it owes the applicant the disclosures, consent and dispute rights that credit-reporting companies have owed consumers since 1970 — obligations that essentially no AI screening tool on the market today is built to satisfy, because none of them were designed against that statute.

Eightfold moved to dismiss on April 20, 2026; the plaintiffs opposed on June 18; Eightfold's reply was filed July 9, and the motion was fully briefed for a hearing on August 4, 2026. As of this writing there's no public ruling. But the theory doesn't need Eightfold specifically to lose for it to matter — it needs a court to accept, even provisionally, that FCRA can reach a hiring-score algorithm at all. Once that door is open, it's open for every vendor whose product produces a number that stands in for a human judgment about a candidate.

Why the pairing matters more than either case alone

Discrimination law and consumer-reporting law are aimed at different failure modes, and that's exactly what makes running them in parallel dangerous for a vendor. Title VII and the ADEA ask what happened to the outcome — did the tool produce a disparate impact on a protected group, measurable in the aggregate. FCRA and ICRAA ask something almost administrative by comparison: did anyone tell the applicant a report existed, get consent, and give them a way to see and dispute it. A vendor can build a screening tool that turns out to be statistically even-handed across every protected group and still lose on the second theory, because disclosure and process failures don't require proving an unfair outcome — they require proving the process was opaque. Mobley is a bet that the outcome was unfair at scale. Eightfold is a bet that the process was hidden, full stop, regardless of outcome. Together they mean a vendor's exposure doesn't collapse to a single question ("is our model biased?") — it multiplies across two separate bodies of law with two separate ways to win.

For any company buying rather than building AI screening — which is most of the market — the practical questions this raises aren't abstract:

Does the tool disclose to candidates that an automated score exists, independent of whether local law (like NYC's Local Law 144) requires it? Eightfold's plaintiffs say they never knew a score existed until they filed suit.
Who generated the score, and is that party prepared to be named as a defendant separately from the employer — the theory both cases are testing is that the vendor, not just the employer, answers for it.
Can the vendor produce, in discovery, records of what the tool actually evaluated and how — Workday's own internal bias reports became part of the record here because litigation reached the stage where a court could compel them.
Is the evaluation criteria fixed and consistent across candidates, or does it vary in ways that would be hard to explain to a judge months or years after the fact?

Where a platform like ours fits, honestly

We build AI screening tools ourselves, so we're not neutral commentators here — and we'd rather say plainly what that means than gesture at reassurance we can't back up. On our platform, every candidate who goes through an AI screening interview is scored against the same defined dimensions — technical, communication and cultural fit, each out of 100 — against a configurable passing threshold set by the employer, and candidates can complete that screening on their own schedule without a human coordinating it. That's a description of a mechanism, not a claim about fairness or an outcome: a consistent rubric applied uniformly is a precondition for defensibility, not proof of it, and we're not going to dress it up as more than that.

What these two cases actually surface is a harder, unglamorous set of engineering and process commitments — durable records of what a tool was asked to evaluate and when, disclosure that reaches every path a candidate might take through the product, not just the default one, and retention that survives a retry rather than being silently overwritten. Those are the kinds of things a plaintiff's discovery request goes looking for, and "we'll say we have it once we actually do" is a slower path to a marketing line than most vendors are comfortable with. It's also the only version of that line a court will find credible if it's ever tested.

The takeaway for HR teams evaluating vendors right now

Neither case has reached a final ruling, and neither should be read as settled law. But the questions being litigated are ones any HR or legal team can ask a vendor today, before a court forces the answer out of one in discovery: What does your tool disclose to candidates, and when? What happens to that disclosure across every entry point into your product, not just the main one? If a regulator or a plaintiff's lawyer asked for a record of what a specific candidate's screening evaluated, could you produce one — and would it still say the same thing a year later? The vendors who can answer those questions before they're asked in a deposition are the ones this wave of litigation is quietly sorting from the ones who can't.

Sources: Lawyer Monthly, "Workday Faces Class Bid Over AI Hiring Bias Claims" (Sept. 2026); FindLaw, Mobley v. Workday Inc. docket; Civil Rights Litigation Clearinghouse, Mobley v. Workday, Inc. case page; National Law Review / Epstein Becker Green, "AI Hiring Tools and Consumer Reports: Understanding the Eightfold Litigation" (also syndicated at workforcebulletin.com); ZwillGen, "Plaintiffs Allege AI Hiring Tools Violate FCRA"; Kistler et al. v. Eightfold AI Inc., complaint (classaction.org).