Meta sued by 26 ex-workers over AI-driven layoff scoring

The interesting thing about this filing is not the layoff itself, it is which employer inputs the plaintiffs put on trial. Twenty-six former Meta workers, all filing anonymously in federal court in Oakland, told The Verge and others that Meta's own AI-flavored performance stack, including performance ratings, calibration scores, productivity metrics, "AI-native" ratings, and AI-token consumption, drove who was selected in the May round, and that those signals structurally under-count anyone who was out on protected leave.

According to reporting relayed by U.S. News and Reuters, all 26 plaintiffs had taken or requested leave under the Family and Medical Leave Act or related state laws within the prior 24 months. They were notified in May that their jobs would end starting July 22, and they are asking the court for a preliminary ruling blocking those terminations while they pursue their individual claims in private arbitration. The complaint also invokes recently adopted California and New York City rules that require employers to test AI evaluation tools for bias, and says Meta did not.

Why this matters if you do not work at Meta: this is a prominent test of the "an AI helped us but humans decided" line that big employers have been holding since generative tools moved into HR workflows. A Meta spokesperson said "workforce management and organizational decisions were and are made by people, not AI" and called the claims meritless. The court will get to look at whether a human sign-off on a list ranked by AI-token dashboards, keystroke data, and internal tools referred to as "Metamate" and employee-trained "second-brain" agents actually counts as a human decision.

The honest caveat is that this is a complaint, not evidence. Nothing has been tested in court, the plaintiffs are anonymous, and the descriptions of Meta's internal tooling come from their side. The reporting does not tell us how heavily any single signal was weighted, whether Meta ran a bias audit against the California or New York City rules, or whether the class will grow beyond the 26 filers.

The part worth watching, whoever wins, is discovery. If a court forces Meta to show what actually feeds these ranking systems, every other large employer running an "AI-native" score into a layoff decision will get a much clearer picture of what a plaintiff can subpoena, and that ripples well outside Meta.