AI isn’t thinking. Or at least Meta wants us to believe that.
Former employees are taking the tech giant to court, alleging exactly the opposite: that hidden algorithms ruthlessly sorted the most vulnerable workers out of 8,000 layoffs this May.
They aren’t just claiming bad management. The suit paints a picture of a surveillance state where productivity scores, token usage logs, and an internal AI chatbot named “Metamate” did the dirty work of human resource departments. And according to the complaint, the AI didn’t just miss the mark on fairness—it systematically penalized people for existing as humans with bodies.
The results? Discrimination against pregnant workers, people on medical leave, and those with disabilities.
The Allegations Behind the Algorithm
Twenty-six former staff members are at the center of this litigation. They claim Meta deployed improperly tested AI systems to rank who stayed and who went.
These rankings weren’t based solely on output quality or project importance. Instead, they relied heavily on metrics that ignored context. Missed days for maternity leave? The AI counted it against you. Time off for a known medical condition? Another strike. Reduced hours due to disability-related accommodations? Scored low on productivity.
“This is patently untrue. Fullstop. Workforce management and organizational decisions were andare made by people, not A
The plaintiffs argue these tools failed basic bias screenings required by law in both California and New York City. They claim the system violated federal anti-discrimination statutes and specific local laws governing how technology can assess employment suitability.
Meta’s Denial and the “People Decisions” Defense
Naturally, Meta is pushing back hard.
Andy Stone, a Meta spokesperson, took to X (formerly Twitter) to call out the allegations. The company’s stance is clear: AI did not fire people.
Workforce decisions were, and remain, human choices. That’s the official narrative. But for the 26 plaintiffs, the distinction feels like a semantic shell game when the algorithm did the heavy lifting of identifying “underperformers” based on skewed data.
They argue the gap between human responsibility and automated bias is too wide to ignore.
Why AI Rankings Favor the “Ideal” Employee
To understand how this happens, look at what the AI was measuring. It tracked things like productivity scores and AI token usage. It likely correlated attendance records with performance metrics without weighting for temporary hardships.
The Model Capability Initiative (MCI) is another piece of this puzzle. This internal tool is designed to train AI models by harvesting employee activity data. Employees noticed the tool collecting significantly more data than initially disclosed.
This raises serious privacy concerns, particularly under European data protection laws which Meta is already under scrutiny for violating in separate contexts. If an internal tracker is building profiles of employee health, break habits, and output variability, it is creating a risk pool. And when layoffs hit, the system flags the high-risk variables—illness, family care, disability—as performance deficits.
It is a feedback loop designed to maximize efficiency at the cost of equity.
The Race Against Time
The lawsuit is not just seeking damages. The immediate goal is to halt the upcoming termination of these 26 workers, currently scheduled for July 22.
The plaintiffs are asking a California federal court for a preliminary injunction. They want the clock to stop. Instead of immediate firings, they want to take their case to private arbitration first. It’s a tactical move, perhaps acknowledging the power imbalance, but also buying time and resources.
The Human Cost of Automated HR
Meta announced these layoffs in the wake of massive AI investments. The message was implied: We need your brains, but we are pivoting toward machine intelligence.
Now, those displaced workers say the machines themselves were used to mark them obsolete before a single human manager sat down for a difficult conversation.
The case hinges on a specific legal and technical question: To what extent does Meta stand liable for the output of a
