The Mobley Workday ruling and the new AI hiring vendor liability landscape
The AI hiring vendor liability ruling in Mobley v. Workday, Inc., No. 3:23-cv-00770-RFL (N.D. Cal.), has redrawn the risk map for every employer using algorithmic screening tools. In an August 2024 order on Workday’s motion to dismiss, Judge Rita F. Lin of the Northern District of California allowed certain state and federal discrimination claims under FEHA, the ADA, and the ADEA to proceed directly against Workday as a vendor, treating its artificial intelligence systems as a potential agent that can shape employment decisions at scale. Other theories, including some federal Title VII claims, were narrowed or dismissed at this stage, and Workday continues to contest liability. For senior talent leaders, the message is that the old assumption that only the employer is liable for discriminatory hiring decisions made with third-party tools is now legally unsafe, even though the ultimate merits of the case remain unresolved.
Court filings in the Mobley Workday case, including the First Amended Complaint and related briefing, allege that Workday’s hiring tools and related systems helped reject roughly 1.1 billion applications during the relevant period (see, for example, ECF No. 37, ¶¶ 3–5, Mobley v. Workday, Inc., No. 3:23-cv-00770-RFL (N.D. Cal.)). Those figures are plaintiff allegations, not findings of fact, but they illustrate the potential impact of algorithmic selection decisions on every protected group in the labor market. When a screening tool touches that scale of employment data, even a modest disparate impact in the selection rate for Black applicants or older candidates can translate into thousands of alleged adverse impact events and discrimination claims. Under this AI hiring vendor liability ruling, a candidate can now argue that the vendor’s algorithmic decision-making is itself an employment practice subject to civil rights law, not just a neutral software product, while vendors will argue that they lack the direct control over hiring outcomes that traditional employers exercise.
The court’s reasoning relied on an agency theory that treats the vendor as an agent of the employer for purposes of an employment decision, at least at the pleading stage, which opens the door for plaintiffs to sue in the vendor’s home state even when the employer and candidate are elsewhere. That shift matters for every employer liable under Title VII and comparable state laws, because it encourages plaintiffs’ attorneys to target deep-pocket vendors and to frame allegedly discriminatory hiring as a joint enterprise between employer and vendor. At the same time, vendors are likely to emphasize contract allocation of responsibility, limits on their authority to make final hiring decisions, and the role of human review to argue against a finding of agency. For HR and TA leaders, the practical takeaway is blunt: if your hiring tools rely on opaque artificial intelligence models, you now share the litigation spotlight with your vendors, and your contracts, bias audits, and documentation of how tools are actually used must catch up fast.
What Mobley changes for HR interviews, screening workflows, and vendor contracts
For HR job interviews, the AI hiring vendor liability ruling turns your early-stage screening workflows into potential exhibits in future discrimination claims. When Workday or any similar vendor provides automated tools that rate individuals before they ever reach a recruiter, those algorithmic decisions become part of the employment record and can be challenged as discriminatory hiring if they depress the selection rate for any protected group. Plaintiffs will point to patterns of adverse impact, while vendors and employers may argue that human decision makers retain ultimate discretion. That is why every employer should now treat vendor liability as a core element of its legal risk register, not a boilerplate clause buried in procurement files.
Under Mobley, plaintiffs can argue that a third-party vendor’s systems are effectively making or heavily influencing hiring decisions, which means the vendor and the employer can both be held responsible for alleged adverse impact under Title VII and state civil rights statutes if agency and control are proven. This dual exposure raises the stakes for bias audits, because a superficial fairness check will not withstand scrutiny if Black applicants or disabled candidates show a statistically significant disparate impact in downstream employment outcomes. In practice, that means you must be able to trace each employment decision from the initial screening tool through HR interviews and final selection, and explain how artificial intelligence scores were used, overridden, or ignored, supported by contemporaneous records rather than after-the-fact explanations.
The discovery phase in Mobley Workday is already a cautionary tale for every AI hiring vendor, as commentators note that Workday’s internal documentation, model behavior, and governance processes will be examined under oath in a way that few HR tech companies have ever faced. TA leaders should study this moment as closely as they study any new sourcing channel, and they should use it to renegotiate contracts with sharper indemnification, stronger data access rights, and explicit obligations for vendors to cooperate with independent bias audits and government inquiries. A practical checklist for those negotiations includes: (1) clear allocation of responsibility for disparate impact analysis and remediation; (2) guaranteed access to model outputs, training data documentation, and audit logs sufficient to respond to EEOC or state agency requests; (3) cooperation clauses for investigations and civil rights litigation; (4) notice and cure periods for material changes to algorithms that could affect selection rates; and (5) indemnification tied to vendor-controlled automated decision-making and failure to follow agreed fairness standards. One example clause: “Vendor shall, upon Employer’s written request, provide all documentation reasonably necessary to evaluate potential discriminatory impact of Vendor’s automated screening tools, including model documentation, configuration settings, and audit reports, and shall cooperate in good faith with any governmental investigation or civil rights proceeding arising from Vendor’s systems.”
A practical framework for researching vendors and preparing candidates in the post-Mobley era
Researching the company has always been a staple of preparing for HR job interviews, but the AI hiring vendor liability ruling means candidates and employers must now research the vendor ecosystem with equal rigor. As a Head of Talent Acquisition, you should brief candidates that early screening may involve artificial intelligence tools and explain how those tools fit into your structured interviewing process, because transparency can reduce anxiety and improve candidate experience. At the same time, your team must interrogate every vendor about their bias audits, their documented selection rate patterns across protected groups, and their readiness to support you if an employment decision is challenged, including their willingness to share data and participate in joint responses to regulators.
A practical framework starts with four questions for any AI hiring tools provider: what data their systems use, how their algorithmic decision-making influences hiring decisions, how they monitor disparate impact and adverse impact, and how they will share the underlying data if the EEOC or a state agency opens an investigation. You should insist on contract language that defines when the vendor is an agent, clarifies when the employer is liable, and sets out joint responsibilities for responding to discrimination claims involving Black applicants or any other protected group. To operationalize this, build a short audit template you can apply to every vendor: (1) document the specific employment decisions the tool influences and the degree of automation; (2) record selection rates by race, gender, age, and disability status where legally permissible; (3) compare those rates to four-fifths rule thresholds and internal benchmarks over time; (4) capture how often human reviewers override automated scores and in which direction; and (5) log remediation steps when adverse impact appears, including model changes, policy updates, and retraining of interviewers.
Finally, use this moment to tighten your own HR interview playbook so that human decision makers can counterbalance flawed algorithmic screening rather than rubber-stamp it. That means training interviewers to question automated scores, documenting why a candidate advances or stalls, and aligning every interview question with a competency model that can withstand EEOC scrutiny if a civil rights complaint arises. To make this concrete, create a short vendor-audit report outline that summarizes tool purpose, data sources, governance owners, key fairness metrics, and recent changes, and keep it alongside your interview guides. If you want to see how upstream choices shape downstream interviews, this guide on why your job description is your worst interview tool shows how even job ads can create hidden bias long before an AI system or recruiter touches a résumé.