Lawsuit Challenges Meta’s Use of Artificial Intelligence to Make Layoff Decisions

A novel lawsuit, filed by 26 former Meta employees in the U.S. District Court for the Northern District of California, alleges that Meta used artificial intelligence (AI) decision-making tools to select employees for a mass layoff, resulting in disabled employees and employees on protected leave being disproportionately selected. Recently, Meta won the first battle of the suit when U.S. District Judge William H. Orrick denied the plaintiffs’ motion for a temporary restraining order, allowing Meta to proceed with the layoffs. Despite this initial win, the merits of the suit are yet to be decided. As companies rely on AI to make the workplace more efficient, the suit highlights the risks associated with AI-assisted employment decisions.

The lawsuit, filed on July 13, 2026, stems from a May 2026 reduction in force in which Meta announced it would lay off about 8,000 employees (10% of its workforce). The suit alleges Meta used a “constellation of internal artificial intelligence systems,” including “Metamate,” an employee-trained “second-brain” agent to assist in selecting employees for the layoff. The plaintiffs allege that Meta’s AI tools tracked, scored, and ranked employee activity and productivity data, such as keystrokes, emails, documents, and browser history. Meta also tracked and ranked AI token usage (which indicates how much employees use AI tools).

The plaintiffs allege that Meta did not suspend these factors for employees on protected leave or have the AI tools account for medical conditions or family care. According to the plaintiffs, use of the AI tools had a disparate impact on workers whose productivity and AI usage scores suffered because they were on protected leave or were disabled. Meta denies the allegations and asserts the decisions were made by “human business leaders.”

The plaintiffs are current and former Meta employees located in California; Illinois; Washington; New York; Washington, D.C.; Pennsylvania; and Florida. Despite their varied locations, the lawsuit alleges the plaintiffs share one common factor: within the 24 months preceding the reduction in force, they each took, requested, or were approved to take statutorily protected leave; attempted to take protected leave and suffered interference; or requested or received a reasonable accommodation for a disability. The plaintiffs allege that they were terminated after requesting or taking bereavement leave, paternity leave, Family and Medical Leave Act (FMLA) leave, medical leave, or after seeking a reasonable accommodation for a disability.

The plaintiffs brought claims under the FMLA, Title VII, the Pregnant Workers Fairness Act, the Americans with Disabilities Act, and their individual state civil rights laws. Under the FMLA, employers are prohibited from using protected leave as a “negative factor” in any employment decision, including reductions in force. Title VII also requires that facially neutral selection criteria not create unjustified disparate impacts among employees. Additionally, California’s Fair Employment and Housing Act includes newly effective regulations that forbid the use of automated decision systems that produce disparate impact discrimination on the basis of disability or sex.

Meta, like many companies, requires its employees to execute arbitration agreements. This means the merits of the plaintiffs’ cases will be determined through individual, confidential arbitration hearings. The current suit seeks an injunction to block the plaintiffs’ termination while arbitration is pending. The court denied the plaintiffs’ temporary restraining order on July 17, 2026, and a hearing on the plaintiffs’ motion for preliminary injunction has been set for August 24, 2026.

Regardless of the ultimate result, the lawsuit highlights some potential dangers of using AI in employment decision-making. AI tools, on their face, appear to provide fair and neutral decisions because they rely on data points, rather than human decision-making. However, the quality of data being used and how it is used should always be considered. Under-scrutinized data points can become “proxy variables” that disadvantage protected workers and result in discriminatory effects.

Before deploying AI tools for critical employment decisions such as hiring and firing, employers should strongly evaluate the data being used and come up with an effective plan for bias testing their data. Some laws, such as New York City’s Local Law 144, already require employers to conduct a bias audit if they are using AI for an automated employment decision. Proposed legislation in California has a similar requirement (the Automated Decisions Safety Act (AB 1018)). Even employers that do not conduct business in a jurisdiction requiring bias audits can still be subject to disparate impact liability under federal and state anti-discrimination laws and should take steps to mitigate their risk.