Twenty-six Meta workers say the company's layoff machine punished people for taking protected leave. Meta says people made the calls, not AI. That difference is the case.
The lawsuit, filed Monday in federal court in Oakland and assigned to Judge William Orrick, puts a hard question in front of Meta: when a company uses AI scores to rank workers before a mass layoff, who is really making the decision?
According to reporting from CNBC, AP and The Guardian, the 26 plaintiffs say Meta leaned on internal systems including Metamate, employee-trained agents and productivity tools that tracked activity such as keystrokes, emails and browsing data before layoffs that began on May 20, 2026. The cuts affected roughly 8,000 workers, about 10% of the company. That is not a small HR experiment. It is a mass layoff with software sitting close to the blade.
Here's the actual claim. Each plaintiff had taken, requested or been approved for medical, family, disability, bereavement, pregnancy or parental leave within the past two years. They say that timing wasn't a coincidence. If a score rewards screen activity, AI token use, messages sent or documents touched, a person on protected leave starts behind everyone still at the keyboard. You don't need a hostile manager for that to become discrimination. You need a metric that can't see the law.
The workers live in California, New York, Florida, Illinois, Pennsylvania, Washington and Washington, D.C., and they filed anonymously. Their lawyers, including Andrea Mazingo of Lumen Law Center, argue that Meta failed to account for protected leave before using automated systems in a decision that affected people's jobs. The complaint points to federal protections under the Family and Medical Leave Act, the Americans with Disabilities Act and Title VII, and also cites rules such as New York City's Local Law 144, which requires a bias audit before certain automated employment decision tools are used.
Meta isn't conceding an inch. A company spokesperson told reporters that workforce management and organizational decisions were made by people, not AI. That is a clean denial. It is also exactly what discovery will test. Did a manager use real judgment on each name, or did a manager simply approve a ranking already shaped by software? Those are different acts. Courts tend to care about the difference.
The leave problem is the whole case
CNBC and AP both describe this as one of the first major lawsuits to challenge a Big Tech company's use of AI in layoff selection, rather than hiring or promotion. That shift matters because layoffs are where companies often move quickly, compress review steps and hide behind business necessity. Frankly, that is where sloppy systems can do the most damage.
The gender claim is not abstract either. Women are more likely to take pregnancy and caregiving leave, and the complaint argues that a system treating absence as lower output would fall harder on them even without a manager writing down anything discriminatory. The plaintiffs also say workers with disabilities or medical accommodations were penalized when their work patterns did not match the productivity signals Meta's systems preferred.
That is the danger with workplace AI. The bias does not always announce itself. It can sit inside a clean-looking dashboard, dressed up as productivity, adoption or efficiency. If the input is time at work and the missing context is legally protected leave, the output is already tilted before any human opens the spreadsheet.
Meta will argue the final decision stayed with people. It has to. Every employer using AI in personnel decisions now wants the same defense: the tool advised, the human decided. But if the human decision is built on a score that already punished protected leave, that defense gets thinner. You cannot wash a bad metric clean by asking a manager to glance at it.
Discovery is the part to watch
The plaintiffs are not only asking for damages. According to AP and The Guardian, they are seeking emergency relief to stop or delay their terminations pending arbitration, along with an independent review of the layoff process. Some of the workers are still employed but face separation later in July. That makes the case current, not historical.
The discovery fight could become more important than the first ruling. If Meta has to explain how Metamate, employee agents, activity dashboards or AI adoption scores influenced the layoff list, other companies will read every line. Amazon, Google, Microsoft, Salesforce, you name it, all have been pushing AI deeper into internal operations while also cutting or reshaping teams. None of them wants a court file showing how these systems actually rank workers when jobs are on the line.
There is a blunt lesson here for any company using AI in workforce decisions. If your system cannot distinguish low output from protected leave, it is not measuring performance. It is measuring presence. That may be convenient for managers trying to move fast, but convenience is not a defense under the ADA or the FMLA.
The case still has to be proved. Meta may show that managers made independent decisions and that the alleged AI scores did not drive the outcome. But the lawsuit has already put a marker down. If companies want to use automated tools to help decide who keeps a job, they will have to show what the tool measured, what it ignored and who had the courage to overrule it.
Also read: CoreWeave's Long-Term Contracts Never Hedged Against Memory Chip Prices • Hinge's Founder Just Bet $18 Million That Swiping Is Broken • OpenAI Is Building a Moveable Screen-Free Speaker as Its First Hardware Product
