Meta workers say its AI quietly decided who lost a job in mass layoffs

Twenty-six current and former Meta employees sued the company Monday, alleging its AI systems, including the Metamate large language model assistant and productivity tracking tools, determined who was laid off in a mass cut of 8,000 workers. The plaintiffs claim Meta used automated scores that penalized workers on protected leave, violating disability and family leave laws, and failed to conduct required bias audits. Meta denies the allegations, saying all workforce decisions were made by people, not AI.

Twenty-six current and former Meta employees are suing the company, claiming its AI systems, not its managers, effectively picked who got cut in a layoff of roughly 8,000 people. The lawsuit, filed Monday in federal court in Oakland and assigned to Judge William Orrick, argues that Meta leaned on a cluster of internal AI tools to score and rank workers ahead of the layoffs that began May 20, 2026. Those tools included Metamate, Meta's internal large language model assistant, an employee-trained "second brain" that tracked workers' documents and communications, and a productivity score built from keystrokes, screen activity, emails and browser history, according to the complaint reported by CNBC and the U.S. News wire. Here's the actual claim. Each of the 26 plaintiffs had taken, requested, or been approved for medical, family, disability, or pregnancy-related leave within the past two years. They say that timing wasn't a coincidence. The complaint alleges that scores tied to output and activity data, by design, cannot be accumulated by someone who was out on protected leave or whose work was slowed by a disability, and that Meta then used those same depressed scores as a factor in deciding who to let go. Plaintiffs live in six states, California, New York, Florida, Illinois, Pennsylvania and Washington, plus Washington, D.C. All filed anonymously. Their attorneys, including Andrea Mazingo of Lumen Law Center, argue the case reaches beyond any one layoff. They say Meta never ran the bias testing that California and New York City now require before employers can lean on automated tools for decisions like this one. New York's Local Law 144 specifically requires a bias audit of automated employment decision tools before an employer uses them, and it's the law plaintiffs point to directly. Meta isn't conceding an inch. A company spokesperson said workforce management and organizational decisions "were and are made by people, not AI." That's a clean, one-line denial. It also happens to be the exact question this case will have to answer. Did a human review each name on the list with real judgment, or did a human just approve a ranking a machine already produced? Those are very different things under the ADA and the Family and Medical Leave Act. Juries tend to notice the difference. A test case for the industry Reporting from CNBC and U.S. News describes this as the first major lawsuit to challenge a Big Tech company's use of AI specifically in how it selected people for layoffs, rather than in hiring or promotion, where algorithmic bias claims have circulated for years. The complaint also raises a gender angle. Women disproportionately take pregnancy and caregiving leave, attorneys argue, and a scoring system that records absence as reduced output falls harder on them by design, whether or not that was the intent. Every large tech company now runs performance systems layered with dashboards, usage metrics and AI-assisted reviews. That's not new. Meta is simply the first to be sued over what happens when that layer gets applied to a mass layoff instead of a routine review cycle. That's the shift here. Amazon and Google have both leaned harder on data-driven performance tools during their own recent restructuring pushes, and neither has faced a suit that names its systems this specifically. That won't last. Plaintiffs' lawyers in employment cases read each other's complaints closely, and a filing this detailed, naming the exact tools, the exact data sources, the exact legal theory, is effectively a template. What discovery could expose Frankly, the discovery process is where this case gets interesting, win or lose. If Meta actually has to hand over how Metamate's scores were calculated and who signed off on the final layoff list, that record becomes public in a way corporate AI vendors have mostly avoided so far. Every company running AI-assisted headcount decisions right now is watching to see how much of that model gets forced into daylight. Meta's layoffs took effect starting May 20, with terminations reportedly running through late July. The lawsuit doesn't ask a court to reverse them. It asks whether the company can keep building performance and productivity scores on top of systems that quietly can't be earned while someone is legally out sick, on leave, or caring for a newborn, and then act as if the score alone told the whole story. Also read: CoreWeave's Long-Term Contracts Never Hedged Against Memory Chip Prices https://startupfortune.com/coreweaves-long-term-contracts-never-hedged-against-memory-chip-prices/ • Hinge's Founder Just Bet $18 Million That Swiping Is Broken https://startupfortune.com/hinges-founder-just-bet-18-million-that-swiping-is-broken/ • OpenAI Is Building a Moveable Screen-Free Speaker as Its First Hardware Product https://startupfortune.com/openai-is-building-a-moveable-screen-free-speaker-as-its-first-hardware-product/