Meta employees sue, allege AI tools scored parental and medical leave against them in 8,000-person layoff

The core allegation

Twenty-six current and former Meta employees filed a 71-page complaint on Monday in the US District Court for the Northern District of California. They allege the company used a "constellation of internal AI systems" to score, rank, and select the roughly 8,000 workers (about 10 percent of the global workforce) who were laid off in the restructuring announced in spring 2026. The lawsuit claims that rather than relying on the considered judgement of managers who knew the work, Meta fed performance ratings, calibration scores, and productivity data into automated tools that determined the layoff list.

Meta did not pause the system for the individualized, leave- and accommodation-neutral review that the law requires.

How the AI systems worked

According to the complaint, Meta combined data from Metamate (its internal large-language-model assistant) with personal agents trained on employee communications and documents. These were supplemented by performance-evaluation systems, AI-adoption rankings, and dashboards displaying individual token consumption. Token counting estimates how much a person uses a chatbot or generative service, but the plaintiffs argue it says nothing about work quality and depends heavily on role, project assignment, and time actually spent on the job. Starting in May 2024, Meta also deployed a new piece of software called the Model Capability Initiative, designed to log every employee action and harvest data to train future AI models.

The ranking is devalued if an employee has taken parental or medical leave, or has requested a disability-related accommodation.

Specific cases cited

Among the examples presented to the court is a female scientist who was on authorised pre-birth leave and received her redundancy notice just two days before giving birth. Another worker, an engineer, states that his performance score dropped because of time away caused by an injury. A third, a manager on medical leave, says he was dismissed only 16 days after the leave began. The plaintiffs argue that protected absences were transformed into a negative factor in the automated evaluation, creating a discriminatory outcome against people with disabilities, those on family leave, and those recovering from illness.

Meta's response and the broader context

A Meta spokesperson told CNBC and other outlets that the claims lack merit and are not based on facts. "Workforce management and organizational decisions were and are made by people, not AI," the company stated. The layoffs are part of a multi-year restructuring that began in 2022: 11,000 positions were eliminated that year, another 10,000 in 2023, and 8,000 more in April 2025, bringing the total cuts to roughly 29,000 roles. The reductions are funding an aggressive AI infrastructure push, with Meta planning to invest up to 145 billion dollars in AI this year, nearly double the previous year's figure.

Workforce management and organizational decisions were and are made by people, not AI.

What the plaintiffs want

The workers are asking the judge to halt the layoffs while their claims are examined through individual arbitration. They also seek reinstatement to their positions, back pay, compensation for lost stock, benefits, and other damages. The case is at an early stage and the accusations have not been proven in court. Several US states, including California, Colorado, and Illinois, have recently passed laws or regulations aimed at protecting workers from AI-related bias and so-called automated decision systems, setting the lawsuit inside a growing regulatory debate.