Employment & Labor Law

Defending the Algorithm™ Employment Law Series | Edition 1

Welcome back to Defending the Algorithm™ - a LinkedIn newsletter from Pittsburgh law firm, Houston Harbaugh, PC, helping defense attorneys, insurance professionals, employment lawyers, corporate counsel and clients to navigate the intersection between artificial intelligence and the law. This newsletter series is specifically targeted to the defense of employment law claims against employers and was written and edited with assistance from Claude Opus 4.8 from Anthropic and Google Gemini 3.0 Pro and with research confirmation by Westlaw Advantage AI with Co-Counsel. It is a companion to our podcast and blog series, available at: Defending the Algorithm™ created by Henry M. Sneath, Esq.

AI tools used in research: Claude (Anthropic) and Westlaw Precision AI. All content was reviewed and verified by the authors.

The AI Vendor as AI Agent: Why “We Just Bought the Software” No Longer Works

In this Edition #1, we take a defense-side overview look at Mobley v. Workday as likely the most consequential AI-and-employment case so far of 2026. In our next edition, we will focus on the brewing discovery battle in Mobley given that Plaintiffs have requested production of the blackbox structure of the Workday AI-based screening tools, including their Candidate Skills Match and HiredScore Spotlight products. Mobley seeks production of the Workday algorithmic bias training and testing data, algorithmic code, methodology, results and communications including the Python Code and .py files in plain text which are used to give a computer instructions and the desired calculations for it to perform. This is the heart of the algorithmic method and is the guts of any AI program. Workday has raised attorney client privilege objections to these discovery requests and briefing is being completed before a hearing. More on that to come.

The Case Against Workday

Derek Mobley applied to more than one hundred jobs through Workday's hiring platform. He was rejected from every one. Several of the rejection emails arrived in the middle of the night, strongly suggesting that no human had reviewed his application before it was rejected. Mr. Mobley is African-American, over forty, and has documented anxiety and depression. He sued, but he did not sue the dozens of employers with the open positions. He instead sued Workday, the vendor whose AI screening system, he alleges, discriminated against him before any of those employers ever saw his resume.

Earlier this year, Judge Rita F. Lin of the Northern District of California formally authorized a nationwide collective action of Mr. Mobley’s Age Discrimination in Employment Act claim. The case is now proceeding through discovery. A merits ruling is plausible this year.

For employers and the defense counsel who represent them, Mobley v. Workday, 3:23-cv-00770-RFL (N.D. Cal.), is perhaps the most consequential AI-and-employment case currently pending in the United States because of the novel theory advanced, and to this point accepted, by the court.

The Novel Holdings That Affect Employers

Mr. Mobley brought discrimination claims alleging disparate impact on the basis of age, race, and sex under Title VII, disability under the ADAAA, age under the ADEA. Workday contended that it could not be held liable under these employment statutes, asserting that it does not qualify as an "employer" with respect to job applicants. The Court nonetheless allowed the claims to move forward, finding that the complaint sufficiently alleged that Workday acted as an agent of its client-employers — a role that brings it within the definition of "employer" under Title VII, the ADEA, and the ADA. The Court subsequently granted preliminary certification of a collective action on the ADEA claim on May 16, 2025. The court reasoned that when a vendor’s AI tool performs a traditional hiring function (screening, ranking, scoring, recommending, or rejecting), the vendor is acting as the agent of its employer-customers, and federal anti-discrimination statutes reach agents.

As AI usage becomes more prevalent in both the workplace and the hiring process, the Mobley case may become a model for similar litigation in the absence of federal or state legislation governing the use of AI in the hiring process.

Why This Matters for the Defense of These Claims by Employers

Pre-Mobley, an employer facing a hiring-discrimination charge could plausibly argue that any flaw was in the vendor’s system, not the employer’s decision-making. The court’s adoption of the agent theory turns that argument inside out. If the vendor is your agent, the vendor’s flaws are your flaws. The employer’s name is on the rejection notice, not the vendor.

Three operational consequences follow.

First, vendor indemnification clauses are usually too thin to matter. Many current enterprise AI subscription contracts cap vendor liability at the prior twelve months of subscription fees. Such caps were negotiated when the technology felt low-risk, and the legal theory advanced in Mobley had not yet matured. But now, the realistic exposure on a certified ADEA collective dwarfs any subscription-fee cap.

Second, disparate impact survived; intent did not. The Mobley court dismissed the intentional-discrimination theory, but let the disparate impact theory proceed. This is dangerous for employers because disparate impact does not require proof of intent — only statistical disparity in outcomes. The four-fifths rule remains the EEOC’s working benchmark, and AI tools producing a selection rate for any protected class below eighty percent of the highest-selected group will draw scrutiny.

Third, the “black box” defense is collapsing. A 2024 University of Washington study of three commercial large-language-model resume screeners, run on more than five hundred real resumes, found systematic preferences favoring white-associated names over Black-associated names a substantial majority of the time. Those findings are now part of the public record, and courts, legislators, and regulators routinely draw on them.

State Legislatures Are Acting

Mobley is not the only pressure point. State legislatures have already undertaken measures to codify protections against workplace AI discrimination. Effective January 1, 2026, Illinois has amended its Human Rights Act to prohibit AI-driven discrimination in hiring, promotion, discipline, and discharge. It also forbids the use of zip codes as proxies for protected classes. Outcome controls; intent is irrelevant.

California’s Civil Rights Department regulations on Automated-Decision Systems took effect October 1, 2025. The California Privacy Protection Agency’s separate automated decision-making technology (ADMT) regulations under the CCPA require risk assessments, opt-out rights, and pre-use notices for ADMT used in significant employment decisions, with full compliance required by January 1, 2027.

The Colorado AI Act has sparked significant debate since its enactment in 2024, prompting numerous proposed amendments from both consumer advocates and industry stakeholders over the following two years. Like its predecessor, the revised law focuses on the use of automated decision-making technology and AI systems ("ADMT") in contexts involving high-stakes consumer decisions. The law defines ADMT broadly to include systems that process personal data and produce outputs — such as predictions, recommendations, classifications, rankings, or scores — that inform decisions about individuals. It applies to ADMT used in "consequential decisions," a term covering determinations that can materially affect a person's life. This includes decisions related to employment and compensation, educational access, housing eligibility, financial and lending services, insurance, healthcare, and essential government services. Any use of ADMT that bears on these areas falls within the scope of the Act.

For most employers, the combination of Mobley-type cases and new state regulatory schemes will quickly impose significant new burdens on the use of AI decision-making in the workplace. Employers must take proactive steps now to ensure compliance.

Four Defense Steps to Take Right Now

1.   Inventory every AI tool that touches an employment decision. Hiring, performance, scheduling, discipline, layoff selection. The first line of defense is knowing what tools you actually deploy. If you are not intimately familiar with the AI tools you use in the employment decision-making process, you will have a very difficult time defending those decisions when challenged.

2.   Demand the vendor’s bias-audit data and methodology. If the vendor will not produce it, that is itself a risk indicator. NYC Local Law 144 already requires annual independent bias audits for Automated Employment Decision Tools used on NYC residents; employers should treat that standard as a baseline regardless of jurisdiction.

3.   Build a meaningful human-review checkpoint. California’s ADMT regulations explicitly exempt processes where a human decisionmaker knows how to interpret the AI output, considers other relevant information, and has authority to override AI decisions. Construct the checkpoint, train the reviewer, and thoroughly document the operation.

4.   Re-examine vendor indemnification at every renewal. Subscription-fee caps will not come close to covering a certified collective action. Push for substantive, comprehensive indemnification from your AI vendors. Whenever possible, have legal personnel review AI vendor renewals.

What Comes Next

Mobley will undoubtedly reshape the way courts handle AI-related workplace discrimination allegations. The moment for employers to adapt and proactively strengthen their AI-related processes is now. We are watching Mobley, and other AI-and-employment related cases and developments closely on behalf of our clients, and we will continue to report.

This is the first edition of a new series within Defending the Algorithm™ devoted to AI in the workplace. Future editions will address discovery of AI infrastructure and architecture in the Mobley case, legal implications in connection with AI hiring and algorithmic layoff selection, potential exposure from utilizing AI meeting tools, as well as legal implications from workforce monitoring.

Connect With Us

Shelly Pagac chairs the Labor and Employment practice at Houston Harbaugh, P.C., in Pittsburgh. Eric Spada, Erica Pietranton and Grant Allison are members of the firm’s Labor and Employment practice.

To discuss AI-related employment exposure or to schedule a vendor-contract review, contact the authors at Houston Harbaugh, P.C., 412-281-5060, or click here to visit our website.

Defending the Algorithm™ is a mark used in interstate commerce by Houston Harbaugh, P.C. (federal registration in progress). This newsletter is informational only and does not constitute legal advice or create an attorney-client relationship.

Sources Verified

Mobley v. Workday, Inc., Case No. 3:23-cv-00770-RFL (N.D. Cal.), May 16, 2025 collective certification order; February 17, 2026 notice authorization order. Illinois Public Act 103-0804 (HB 3773), effective January 1, 2026. California Civil Rights Department Final Regulations on Automated-Decision Systems, effective October 1, 2025. California CPPA ADMT Regulations, compliance required January 1, 2027. Colorado SB 26-189 (signed May 14, 2026; effective January 1, 2027), Colorado SB 24-205 and SB 25B-004; April 27, 2026 federal enforcement-pause order. Executive Order 14365 (December 11, 2025), 90 Fed. Reg. 239 (Dec. 16, 2025). NYC Local Law 144 of 2021 and DCWP Final Rule (April 2023). EEOC Uniform Guidelines on Employee Selection Procedures (four-fifths rule). EEOC v. iTutorGroup settlement (August 2023). ACLU/Public Justice complaint against Intuit and HireVue (March 19, 2025), filed with EEOC and Colorado Civil Rights Division. University of Washington 2024 study on AI hiring bias (Wilson and Caliskan).