AI on the Layoff List: Meta Lawsuit and Workday Rulings Raise the Stakes for Algorithmic Employment Decisions
On July 13, 2026, 26 current and former Meta employees sued the company in federal court in Oakland, California, alleging that Meta used artificial intelligence (AI) systems to select them for layoff in a manner that discriminated against workers with disabilities, workers on protected medical or family leave, and pregnant employees. The Complaint, Does 1–26 v. Meta Platforms, Inc, is the first major challenge to AI-assisted reduction-in-force decisions. It follows a series of rulings in Mobley v. Workday, Inc that expanded potential liability for AI-driven hiring tools, including for the vendors that build them, as we touched on in our previous Client Alert. Employers using automated or AI-assisted tools anywhere in the employment lifecycle should treat these cases as a preview of the claims coming their way.
The Meta Lawsuit
Meta announced in May that it would lay off roughly 8,000 employees, about 10 percent of its workforce, with terminations effective beginning July 22. The plaintiffs allege the termination list was not built through the “judgment of managers who knew the work.” Instead, they claim Meta relied on a “constellation of internal artificial-intelligence systems” to score, rank, and select employees, drawing on keystroke and activity-monitoring data, AI token-usage dashboards, and algorithmically assisted performance rankings.
The theory is disparate impact by proxy. The Complaint alleges that these metrics, by design, cannot be accumulated by an employee on protected medical or family leave, or by one whose output is reduced by a disability. The inputs are facially neutral; the outputs, plaintiffs say, are not. Notably, the Complaint does not allege that Meta fed medical information into an algorithm. It alleges that the algorithm penalized the absence of activity that protected leave necessarily produces.
The 26 anonymous plaintiffs come from six states and the District of Columbia. The particular leave varied among the individual plaintiffs, including pregnancy-related leave, parental leave, and medical, caregiver, or bereavement leave, as well as disability accommodation requests. The claims span the Family and Medical Leave Act (FMLA), the Americans with Disabilities Act (ADA), the Pregnancy Discrimination Act (PDA), the Pregnant Workers Fairness Act (PWFA), California Fair Employment and Housing Act (FEHA) (including its automated-decision regulations), and civil rights or paid leave laws in New York, New York City, Washington, DC, Washington State, Florida, and Illinois. Plaintiffs also allege Meta failed to test its AI systems for bias as required by recently adopted California and New York City laws.
The immediate relief sought is narrow but urgent: a preliminary injunction blocking Meta from completing the layoffs while the plaintiffs pursue their claims in individual arbitration. Plaintiffs contend their arbitration agreements do not reach requests for temporary judicial relief. Meta denies the allegations, stating that workforce decisions “were and are made by people, not AI.”
Mobley v. Workday: The Vendor Is in the Case to Stay
The Meta complaint arrives against the backdrop of Mobley v. Workday, the closely watched challenge to Workday’s AI-powered applicant screening tools, pending before Judge Rita Lin in the Northern District of California. Derek Mobley alleges he was rejected from more than 100 positions at companies using Workday’s platform, at times within minutes or in the middle of the night. In May 2025, the court conditionally certified a nationwide Age Discrimination in Employment Act (ADEA) collective action, noting that Workday estimated roughly 1.1 billion applications were rejected through its tools during the relevant period.
Two recent rulings sharpen the exposure. On June 22, 2026, the court declined to dismiss most of the remaining claims. It held that California’s FEHA can reach non-California resident applicants because Workday designed, developed, and maintains its AI tools at its California headquarters, thus creating a sufficient nexus with California. It also allowed an ADA claim to proceed on the theory that the software screened an applicant out based on proxies for disability, including patterns consistent with medical leave, treatment, and recovery. A race-based disparate impact claim was dismissed on procedural grounds. The court had earlier held that Workday could face liability as its customers’ “agent” because employers plausibly delegated the traditional function of rejecting and advancing candidates to the software.
A May 29, 2026 discovery ruling deserves equal attention. The court denied the plaintiffs’ motion to compel Workday’s bias-testing data, holding that the attorney-client privilege protected it because Workday’s counsel curated the testing and used the results to provide legal advice. For employers deciding how to audit their own AI tools, the lesson is direct: bias testing conducted at the direction of counsel can remain privileged; bias testing run as an ordinary business function likely will not.
What This Means for Employers
Disparate impact remains a live private theory. Although the current administration has directed federal agencies to deprioritize disparate impact enforcement, as detailed in our prior Client Alert, private plaintiffs are under no such instruction, and state statutes, including Michigan’s Elliott-Larsen Civil Rights Act and Persons with Disabilities Civil Rights Act, supply independent avenues. The Meta plaintiffs built their complaint largely on state law.
Productivity metrics can become proxies for protected status. Any selection criterion that protected leave depresses, including activity data, output volume, or AI tool usage, invites the Meta protected leave from the underlying data and run an adverse impact analysis on the selection criteria, ideally under privilege.
Using a vendor’s tool does not shift the risk and may import new risk. The EEOC’s position has long been that employers are responsible for discriminatory outcomes produced by vendor-built tools. Mobley adds a second layer: a tool designed and maintained in California may carry FEHA exposure to workforces far outside that state. Vendor contracts should be revisited for audit rights, bias-testing documentation, and indemnification for algorithmic discrimination claims.
Keep documented human judgment in the decision. Meta’s first line of defense is that people, not AI, made the final calls. That defense is only as good as the record behind it. Employers should be able to show who reviewed algorithm-generated rankings, what discretion they exercised, and where human judgment departed from the machine’s output.
Finally, audit under privilege. The Mobley discovery ruling gives employers a roadmap for testing AI tools for bias without creating a discoverable record: involve counsel at the outset, have counsel direct the scope and methodology, and route results through counsel for legal advice.
Our Labor and Employment group is advising employers on AI governance in hiring, performance management, and reduction-in-force planning. If your organization uses automated tools in any of these areas, we are happy to discuss a privileged review.
Please contact the authors of this Client Alert or your Butzel attorney for more information.
Brett Miller
313.225.5316
millerbr@butzel.com
Michelle Cirino
313.225.5306
cirino@butzel.com
