Companies across industries are rapidly adopting artificial intelligence (“AI”) for hiring, performance evaluation, and workforce decisions. At the same time, the legal risks of using these tools are mounting. Last month, Meta Platforms, Inc. (“Meta”) became the highest-profile company yet to face a federal lawsuit alleging its AI systems produced discriminatory employment outcomes.
The Meta lawsuit is a warning: employers using AI to evaluate employees and candidates must proceed with caution and cannot wait for regulatory clarity to act. Proactive auditing, training, and compliance are essential to the responsible use of such tools and to making defensible workplace decisions.
Overview of the Meta Lawsuit
On July 13, 2026, twenty-six (26) current and former Meta employees filed suit in the Northern District of California, alleging Meta’s internal AI systems selected them for layoff in violation of the Family and Medical Leave Act (“FMLA”), the Americans with Disabilities Act (“ADA”), the Pregnancy Discrimination Act (“PDA”), the Pregnant Workers Fairness Act (“PWFA”), Title VII of the Civil Rights Act of 1964, and several California laws including regulations that extend the state’s Fair Employment and Housing Act (“FEHA”) anti-discrimination framework to “automated-decision systems” used in employment.
The plaintiffs—engineers, managers, researchers, and designers across seven jurisdictions—are among the approximately 8,000 workers (roughly 10% of Meta’s workforce) Meta announced it would cut in May 2026. Each either took protected leave or requested a disability accommodation, including one scientist who received a layoff notice two days before her due date, after her maternity leave had already been approved.
How the AI Allegedly Discriminated
The complaint targets what it calls Meta’s “constellation of internal artificial-intelligence systems,” which allegedly relied on performance ratings, calibration scores, productivity metrics, “AI-native” ratings, and AI-token consumption data to evaluate its employees. The core problem is that employees on protected leave or with disabilities cannot accumulate these metrics. The lawsuit identifies three tools:
- “Metamate,” a large language model that allegedly tracked employee communications and documents;
- A “second brain” system that allegedly monitored workers’ activities; and
- A productivity score allegedly derived from keystrokes, screen content, emails, and browser history.
Critically, Meta did not pause these monitoring systems during legally protected leave, so employees’ scores dropped during absences. The plaintiffs allege that these scores served as inputs for layoff selection, making discrimination inevitable.
The lawsuit also references Meta’s Model Capability Initiative (MCI), a program that—with no option for employees to opt out—captured keystroke data, mouse movements, click locations, and periodic screenshots from employees’ work laptops. The MCI was paused in June 2026 after a security failure exposed private employee conversations and performance data company-wide, confirming both the scope of Meta’s behavioral monitoring and its inadequate data governance.
The Legal Framework: Disparate Impact Applies to AI
The lawsuit relies on the disparate impact doctrine, which is a legal principle that facially neutral employment practices are unlawful if they disproportionately harm a protected group without business justification. In plain terms, a company can be liable even without intending to discriminate if a policy has an unfair effect on certain groups. Although the Trump Administration has sought to limit the doctrine, that policy shift does not affect private lawsuits. Workers remain free to bring disparate-impact claims on their own, and several state laws—including those in New York, California, New Jersey, and Illinois—independently prohibit the same conduct. In short, the disparate impact framework applies with full force to private lawsuits aimed at AI-driven employment decisions.
The Relief Sought and the Court’s Initial Ruling
The plaintiffs sought an emergency order to block layoffs scheduled for July 22, 2026, and to require an independent AI audit. Meta argued the reduction in force was a business-driven reorganization, with selection decisions made by human leaders based on neutral criteria which did not include leave status or history, disability status, accommodation requests, or any other protected characteristic.
On July 17, 2026, U.S. District Judge William Orrick denied the temporary restraining order but found “serious questions going to the merits” of plaintiffs’ claims. The Court set a preliminary injunction hearing for August 24, 2026, though it noted the merits will ultimately be resolved in private arbitration pursuant to the employees’ arbitration agreements. The private arbitration has already commenced.
The Rise of AI Discrimination Cases
The Meta lawsuit does not arise in a vacuum. It is part of a growing wave of AI discrimination cases brought by employees and job candidates across the country.
One of the first of these lawsuits was Mobley v. Workday, Inc., Case No. 3:23-cv-00770-RFL (N.D. Cal.), filed in February 2023. There, a federal court allowed sweeping discrimination claims against Workday’s AI applicant screening system to proceed to discovery. The plaintiff alleges Workday’s AI discriminated on the basis of race, age, and disability. Workday’s platform processes hundreds of millions of applications annually—the company disclosed that it rejected applications numbering in the billions during the relevant period.
Key developments in Mobley underscore the growing legal exposure when companies use AI systems in hiring decisions. The court conditionally certified a nationwide ADEA collective action, rejected Workday’s motions to dismiss, and required Workday to produce EEOC reporting information and demographic data, finding it relevant to Workday’s knowledge of potential disparities in its AI tools.
Other cases are expanding on these legal theories. For example, in Kistler v. Eightfold AI Inc. (N.D. Cal. 2026), plaintiffs are suing an AI hiring vendor under the Fair Credit Reporting Act, targeting the vendor directly rather than the employer, and alleging it collects sensitive and inaccurate information about job applicants to score them from zero to five, which employers then use to sift through applications. Eightfold’s motion to dismiss is currently pending before the court.
In Johnson-Rocha v. Immunovant, Inc. (S.D.N.Y. 2025), a job candidate sued an employer under traditional anti-discrimination statutes for using AI screening tools during the hiring process, alleging those tools filtered out her job applications. A jury trial is scheduled for October 2026, and the employer’s motion for summary judgment is pending.
The trend is clear: AI systems influencing employment decisions expose both vendors and employers to liability—regardless of whether AI makes the “final” decision.
Key Takeaways for Employers
Employers cannot escape discrimination liability by delegating decisions to AI, and relying on such tools even where a human is the decisionmaker comes with risks. Courts are holding that when AI outputs influence hiring, promotion, or termination, the employer bears responsibility for discriminatory impact. Prevention is the best defense.
Employers should take the following steps now to help mitigate the risks and embrace their workplace’s AI usage with confidence:
- Audit AI systems for disparate impact. Evaluate how your organization uses AI in employment decisions. Test any AI tool used in hiring, performance evaluation, or layoff selection for adverse impact on employees with disabilities, those on protected leave, and other protected classes. Conduct a comprehensive risk assessment and establish a remediation roadmap.
- Pause monitoring during protected leave. If AI-driven productivity metrics factor into employment decisions, ensure scoring systems pause during legally protected absences. Failure to do so creates the structural bias alleged in the Meta case.
- Develop a comprehensive AI use policy. Establish a framework that addresses rules of use, transparency and disclosure obligations, data governance and privacy, and management of AI agents. Involve stakeholders in policy development to ensure organizational buy-in.
- Maintain human oversight. AI should inform, not determine, employment decisions. Ensure human decision-makers review and can override AI recommendations with documented rationale. Avoid having your decision-makers fall prey to “cognitive surrender”—the tendency of employees to accept AI outputs uncritically—through training and clear accountability standards.
- Vet your Require transparency about how third-party AI platforms score, rank, or filter candidates. Contractually require bias testing, indemnification, and compliance with state obligations. The Kistler case demonstrates that AI vendors themselves face direct liability—but that does not absolve employers who use their tools.
- Monitor the regulatory landscape and implement training. The Trump Administration has implemented a policy of deregulating AI systems at the federal level, leaving open a void that states are filling. New York, Illinois, Colorado, Connecticut, and Maryland have enacted AI-specific employment laws, and additional states—including Massachusetts and Rhode Island—have pending legislation. Multi-state employers must monitor this evolving landscape. Implement regular training for employees who use or interact with AI systems and establish monitoring and enforcement mechanisms to ensure policy compliance.
The legal landscape is evolving rapidly, but one thing is clear: courts will not let AI serve as a shield for discrimination. Companies deploying these systems without rigorous testing, human oversight, and leave-aware design face significant legal risk. It is vital to seek experienced counsel to establish AI policies and implement compliance guardrails
