When Meta cut thousands of jobs earlier this year, the company framed it as a performance-driven purge — a rational, data-informed thinning of the herd. What a federal lawsuit filed in Oakland, California now alleges is something considerably darker: that the algorithm doing the culling was systematically punishing employees for exercising federally protected rights.

Twenty-six plaintiffs, all former Meta workers, accuse the company of deploying AI-powered software that weighted factors like productivity metrics and AI token usage — measurements that, by their nature, drop to zero when someone is on medical leave, recovering from a disability-related procedure, or caring for a newborn. Under that scoring logic, a software engineer who spent six weeks on chemotherapy would look, on paper, like a chronic underperformer. According to the complaint filed in the U.S. District Court for the Northern District of California, that is precisely what happened.

The lawsuit claims the AI-driven selection process disproportionately swept up workers covered under the Americans with Disabilities Act, the Family and Medical Leave Act, and California's parallel state protections — laws that exist specifically to prevent employers from penalizing workers for medical absences. Whether the discrimination was intentional or baked structurally into the model's design, the plaintiffs argue, makes no legal difference. Disparate impact is disparate impact.

The numbers cited in the complaint are striking. Of the roughly 8,000 employees affected by the layoff wave, the suit alleges a statistically anomalous share were individuals who had taken medical, parental, or family leave in the period the algorithm evaluated. That pattern, the plaintiffs argue, is not coincidence — it is the predictable output of a system that measured presence as a proxy for value.

This is the central problem with automating consequential employment decisions: algorithms do not know why a number is low. They register absence, not context. A model trained to maximize output scores will not distinguish between a low-performer and a new mother on parental leave; both look identical in the data. Designing a layoff system without accounting for that — without flagging or excluding protected leave periods from the productivity calculus — is either negligence or indifference. The lawsuit, in effect, argues it was both.

Meta has not publicly addressed the specific allegations in the complaint. The company's broader posture in 2025 has been to frame its workforce reductions as a deliberate sharpening of focus, with Zuckerberg describing the cuts in internal communications as a move to raise the performance bar across the organization. What those internal communications also revealed — in audio that leaked shortly after Meta's CTO expressed his desire to sue leakers — is that leadership was aware that the cuts would generate legal exposure. The irony of that leak was not lost on the workforce that survived the round.

Legally, the case sits at an unresolved frontier. Courts have dealt with algorithmic hiring discrimination before, but AI-driven termination at this scale — and with this level of documented metric dependency — is relatively new ground. The Equal Employment Opportunity Commission has issued guidance warning employers that using automated tools does not insulate them from discrimination liability, and that facially neutral screening criteria can still violate federal law if they produce discriminatory outcomes. That guidance, however, was advisory. A federal court ruling in this case would carry substantially more weight.

What the lawsuit ultimately reveals is a leadership failure that is structural, not incidental. Any competent employment attorney advising on a mass layoff would flag the need to audit protected-leave status before finalizing a termination list generated by any automated system. The fact that twenty-six people with documented medical and parental leave histories ended up on the cut list — and felt the need to organize and sue — suggests that audit either did not happen or its findings were not acted on. For a company that has made AI the centerpiece of its identity and its pitch to investors, that is a meaningful failure of the thing it claims to do best: build systems that work.