Meta fired roughly 8,000 people using an AI-assisted performance system that, almost by design, punished anyone who had the audacity to take medical or parental leave. Now 26 of those workers are suing the company before their separations go final on July 22, and Meta's response is essentially: the robots didn't do it, people did. Very reassuring.
What the System Actually Did
Here is the setup. According to the lawsuit filed late Monday in federal court in Oakland and reported by CBS News, Meta used a cocktail of internal AI systems, keystroke monitoring, activity tracking, AI token-usage dashboards, and algorithmically assisted performance rankings to decide who was getting cut. The company had already announced in April that 10% of its workforce, around 8,000 jobs, was going.
The problem, as the lawsuit spells out with uncomfortable clarity, is that many of those scores and ratings 'by design, cannot be accumulated by an employee who is on protected medical or family leave, or whose output is reduced by a disability.' You are not generating AI tokens while you are in a hospital bed. You are not logging keystrokes while you are recovering from childbirth. The system did not care. It just saw the zeros.
Meta, per the lawsuit, made no adjustment for any of this. No pause for protected leave. No individualized review. Just the algorithm, humming along, quietly marking down the people who had the gall to get sick or have a baby.
Who Actually Got Caught in This
The 26 plaintiffs are anonymous in the filing, but CBS News reports that about half of them took leave for caregiving or pregnancy-related reasons. Eight are women who took maternity or pregnancy-related leave. Four are men who took parental leave. One woman took leave to care for a family member and then took bereavement leave after that family member died.
Then there is the worker who had a serious health condition approved by Meta's own disability provider. According to the lawsuit, a manager specifically warned this employee that taking leave would result in being selected for layoffs. He was 'discouraged and deterred' from using leave he was legally entitled to. Meta then offered no disability accommodation. The algorithm selected him anyway.
All 26 remain employed for now, with separations set to begin July 22. The lawsuit is explicitly asking for one thing: keep these people employed while the case goes to arbitration. Once those separations are final, the plaintiffs' lawyers note, the damage is irreversible. Health insurance gone during pregnancy or active medical treatment. Unvested equity forfeited. For workers on visas, immigration consequences triggered.
Meta's Explanation Is a Work of Art
Meta issued a statement saying the claims 'lack merit and are not based on facts.' So far, so standard. Then came the good part: 'Workforce management and organizational decisions were and are made by people, not AI.'
Let that one sit for a second. This is the same company that built the AI performance tracking systems. The same company that fed keystroke data and activity monitoring and token-usage dashboards into algorithmic ranking tools. The same company that then used those rankings to execute layoffs across 10% of its workforce. But when a lawsuit shows up, suddenly it was just people making thoughtful decisions all along. Nothing to see here. The robots were merely decorative.
This is becoming a familiar corporate two-step in the AI age: deploy automated systems to make consequential decisions, then claim human accountability when those decisions turn out to be discriminatory. You get the efficiency of the algorithm and the legal cover of blaming a manager somewhere.
The Legal Minefield Meta Just Walked Into
The lawsuit alleges violations of the Family and Medical Leave Act, the Americans with Disabilities Act, the Pregnancy Discrimination Act, and the Pregnant Workers Fairness Act. That is a fairly comprehensive list of laws specifically designed to protect exactly the people Meta's system appears to have targeted.
The case also leans on disparate impact liability, the longstanding civil rights concept rooted in Title VII of the 1964 Civil Rights Act, which holds that a policy can be discriminatory even if it looks neutral on its face, if it disproportionately harms a protected class. The Trump administration has been trying to gut disparate impact enforcement, ordering the EEOC to deprioritize such cases and arguing the doctrine undermines 'meritocracy.' The EEOC has already dropped discrimination cases on behalf of some workers as a result.
But as CBS News points out, private plaintiffs can still bring these cases themselves even if the EEOC won't touch them, and several states have laws that explicitly prohibit disparate impact discrimination. Meta's lawyers built a system that may have been technically legal in some abstract sense while still being straightforwardly discriminatory in practice. The courts get to sort that out now.
Why This Case Is Bigger Than Meta
This lawsuit is not really just about 26 people at one very large technology company. It is about what happens when you hand hiring and firing decisions to systems that were never designed with legal obligations in mind. An AI performance tracker does not know about the FMLA. It just knows you were not logging activity for six weeks.
Every major company is right now either deploying these kinds of tools or thinking hard about deploying them. The cost savings are obvious. The legal exposure is apparently less obvious, or was until Monday. What Meta may have accidentally done is become the test case that forces the entire industry to reckon with what it actually means when an algorithm makes a consequential employment decision about a human being.
The plaintiffs' lawyers put it plainly: the company's 'algorithmically assisted selection process, by systematically recording such absences as reduced performance, falls more heavily on women than on men.' Because women disproportionately take pregnancy and caregiving leave. The algorithm did not know that. Or maybe it just did not care.
The Dingo Take
Meta just spent the last several years telling the world that AI is the future of everything, that its systems are sophisticated and transformative and will change how we work and live. Then it built an AI performance monitoring apparatus, used it to rank employees, used those rankings to fire 8,000 people, and somehow did not account for the fact that people on legally protected medical leave would score badly on a system that measures activity. That is not a glitch. That is the system working exactly as designed, and the design was reckless.
The 'people made these decisions, not AI' defense deserves to be mocked loudly and often. You do not get to build an algorithmic ranking machine, feed it into your layoff process, and then point at a VP somewhere when the results turn out to be discriminatory. That is accountability laundering. It is the corporate version of 'I was just following orders,' except the orders came from a dashboard you commissioned and deployed yourself.
Twenty-six people are trying to hold onto their jobs, their health insurance, their equity, and in some cases their immigration status while a company worth more than a trillion dollars insists its process was perfectly fair. The separations start July 22. Whatever the courts ultimately decide, the actual question this lawsuit is asking is a simple one: are we going to let companies use AI to strip workers of their legal rights while pretending a human made every call? Because if the answer is yes, every protection workers have ever won becomes negotiable the moment someone writes the right algorithm.
