A federal court in Oakland, California has ordered Meta to file a sworn declaration explaining exactly how and why four company-sponsored visa holders were selected for termination, marking the first time the company has been required to formally account for its layoff selection process.
The order stems from a lawsuit filed on 13 July 2026 by 26 current and former Meta employees, who allege the company's May 2026 reduction in force, which reportedly affected around 8,000 workers, was driven by a suite of internal AI systems rather than managerial judgment.
According to the complaint, the selection process relied on an internal large language model called Metamate, alongside keystroke and activity monitoring tools that captured keyboard inputs, screen activity, email usage, and browser history. A separate dashboard tracked each employee's AI token consumption, measuring how actively they used Meta's internal AI tools. These data streams fed into algorithmically generated performance rankings, including an "AI-native" rating.
The core legal claim is that those inputs structurally disadvantaged workers on protected medical, parental, or disability leave, because time away from work meant lower activity scores. Several plaintiffs claim they were on maternity leave, medical leave, or had disability accommodations when they were selected. One plaintiff alleges she was notified of her layoff just two days before giving birth.
US District Judge William Orrick declined to issue a temporary restraining order blocking the broader layoffs, ruling that most financial harms, lost health cover, share awards, could be remedied later through arbitration. But he treated the four visa holders differently, finding that the threatened loss of immigration status was "more than speculation" and could constitute irreparable harm.
Workers on H-1B and similar visas face a 60-day grace period once employment ends to find a new sponsoring employer, after which they risk losing their lawful immigration status in the United States.
Meta has denied that AI made the final layoff decisions. A spokesperson said workforce decisions "were and are made by people, not AI," and called the claims meritless. A preliminary injunction hearing is scheduled for 24 August 2026.
