A lawsuit filed by 26 employees alleges that Meta used a suite of internal artificial intelligence systems to select workers for termination, and that those systems penalized employees who were on protected medical or family leave. Meta denies the claim, stating that workforce decisions were made by people. The case sits at the intersection of employment law, algorithmic management, and a question that is becoming harder to avoid: when AI tools shape a consequential decision, who is actually responsible for the outcome?
The Architecture of an Algorithmic Layoff
According to the complaint, filed in US District Court for the Northern District of California, Meta did not rely on direct managerial judgment to build its termination list. Instead, the lawsuit describes a layered system of internal tools. These included a system referred to internally as “Metamate,” employee-trained “second-brain” agents, keystroke and activity monitoring data, AI token usage dashboards, and algorithmically assisted performance ranking and calibration.
One detail stands out in particular. The lawsuit states that Meta’s internal dashboards classified employees according to how deeply they had adopted AI tools, using categories such as “AI Native,” “AI First,” and “AI Enabled.” In other words, one of the inputs used to evaluate employees was how much they used AI. Workers who scored lower on AI adoption were, by the logic of the system, less valuable.
This is what most coverage of algorithmic management misses. The issue is not simply that software was involved in a decision. The issue is that the criteria embedded in that software carry assumptions about what a productive employee looks like, and those assumptions may not be neutral.
What the System Could Not See
The core legal argument in the lawsuit is straightforward. The metrics used to score employees, including performance ratings, productivity outputs, AI token consumption, and AI adoption stage, are metrics that an employee on approved medical or family leave cannot accumulate. A person recovering from surgery, managing a disability, or on maternity leave is not generating AI tokens. They are not logging keystrokes. Their output metrics are, by definition, lower.
The lawsuit alleges that Meta did not adjust scores to account for protected leave status, did not exclude employees on leave from the selection pool, and did not pause the system to allow for individualized review. The result, the plaintiffs argue, was that employees who exercised legally protected rights were disproportionately selected for termination because the system treated their absence as underperformance.
The complaint includes a specific account: a scientist was selected for termination while on approved pre-birth pregnancy leave, the day before her water broke and two days before she gave birth. Other plaintiffs were allegedly on maternity or paternity leave, medical leave for disabilities, or working under approved remote accommodations when they were selected.
The 26 plaintiffs all requested leave or disability accommodations in the 24 months before being selected. The layoffs, which affect approximately 8,000 employees, were announced even as Meta reported record revenue and committed to spending between $125 billion and $145 billion on artificial intelligence in 2026, more than double its 2025 expenditure. The lawsuit notes that employees questioned why cuts were necessary given that context.
Meta’s Chief People Officer Janelle Gale had communicated to staff that the company would cut approximately 10 percent of employees and pause hiring for around 6,000 open roles as part of an effort to run the company more efficiently.
Meta’s response has been categorical: “Workforce management and organizational decisions were and are made by people, not AI.”
Why This Case Is Larger Than One Company
The lawsuit is described, citing Reuters, as the first against a major US company to challenge the alleged use of AI in conducting layoffs. That framing matters. It signals that a legal framework for algorithmic employment decisions does not yet fully exist, and that courts and regulators are being asked to develop one in real time.
The plaintiffs allege violations of the Family and Medical Leave Act, the Pregnancy Discrimination Act, the Americans with Disabilities Act, and the Pregnant Workers Fairness Act, among others. They also cite California’s Fair Employment and Housing Act, which has been updated to prohibit the use of automated decision systems that produce disparate-impact discrimination on the basis of disability or sex, including pregnancy.
This is the structural question the case forces into the open. Existing anti-discrimination law was written for human decision-makers. When a manager makes a biased decision, there is a chain of accountability. When an algorithm produces a biased outcome because of how it was designed, what it was trained on, or what inputs it was given, the accountability chain becomes harder to trace and easier to obscure.
The plaintiffs are not seeking a class action. Meta’s employment agreements include arbitration clauses that waive class action rights, so each plaintiff must proceed individually. What they are asking for now is an injunction to preserve their employment status while arbitration unfolds, and an independent audit of the entire selection process, including the inputs, weights, and outputs of the algorithmic system.
In Short
This case is not primarily about whether Meta acted in bad faith. It is about whether AI-assisted decision systems can inadvertently encode discrimination by treating absence as failure. When a system scores employees on metrics they cannot accumulate while exercising legal rights, the system does not need to be designed with discriminatory intent to produce discriminatory outcomes. That distinction, between intent and design, is what courts will now have to work through. The answer will shape how every company using algorithmic tools to manage its workforce understands its legal exposure.
Based on reporting from Ars Technica.