The system turned medical vulnerability into a reason to fire
When a corporation delegates the gravity of a layoff to an algorithm, it does not escape moral accountability — it merely obscures it. Twenty-six former Meta employees, let go during an 8,000-person workforce reduction in 2024, are now asking a federal court in Oakland to examine whether the company's AI-driven performance metrics quietly punished workers whose medical leave, disabilities, or pregnancies made their output appear diminished. The case poses a question that will outlast any single verdict: can a system be neutral when the conditions it measures are not?
- Twenty-six anonymous plaintiffs allege that Meta's AI performance tools — including metrics tied to 'AI token usage' — structurally disadvantaged workers who had taken medical leave or had disabilities, making their terminations feel less like decisions and more like outputs.
- Meta insists humans made the final calls, but the lawsuit cuts deeper: if an algorithm determines who appears on the list, the human signature at the bottom may offer less protection than it seems.
- The case lands amid growing alarm from regulators and employment lawyers who warn that AI systems trained on productivity data can silently encode discrimination against protected classes without anyone explicitly choosing to do so.
- If it reaches trial, the lawsuit could force tech companies to audit their workforce algorithms before deployment and demonstrate compliance with anti-discrimination law — or risk becoming the next cautionary precedent.
Meta is facing a federal lawsuit from 26 former employees who allege the company used AI-assisted performance metrics to systematically disadvantage workers with disabilities, serious medical conditions, or pregnancy during mass layoffs earlier this year. Filed in Oakland, California, the complaint centers on a structural problem: any system that measures productivity or platform activity will inevitably flag workers who took protected leave as lower performers — not because they were, but because they were absent for legitimate medical reasons.
The layoffs, which cut roughly 10 percent of Meta's global workforce across multiple rounds beginning in May, affected approximately 8,000 employees. The 26 plaintiffs, drawn from California, New York, Washington D.C., and other states, argue that relying on these metrics as a primary filter violated both federal and state employment protections for workers with disabilities and medical circumstances.
Meta has denied the allegations, stating that workforce decisions were made by people, not machines. But the lawsuit challenges that framing directly — arguing that when an algorithm shapes which employees are surfaced as candidates for elimination, the human who signs off inherits a pre-filtered reality.
The case arrives as AI in human resources faces intensifying scrutiny. Automation in hiring, performance reviews, and workforce planning has grown rapidly, with companies drawn to its promise of speed and apparent objectivity. Critics, however, have long warned that algorithms built on historical output data can quietly encode discrimination against protected groups.
Should the case proceed to trial, it may become one of the most consequential legal tests of AI in employment decisions — potentially requiring companies to audit their systems before deployment and demonstrate that efficiency-driven tools do not come at the expense of workers whose circumstances made lower output unavoidable.
Meta is being sued by 26 former employees who claim the company used artificial intelligence to systematically disadvantage workers with disabilities and those who had taken medical leave during mass layoffs earlier this year. The lawsuit, filed in federal court in Oakland, California, centers on an uncomfortable question: when a company automates the decision about who stays and who goes, who bears the cost of the algorithm's blind spots?
The plaintiffs allege that Meta relied on AI-assisted performance metrics—specifically measuring employee productivity and what the complaint calls "AI token usage"—to determine which workers would be cut. The problem, they argue, is built into the logic itself. Employees who had taken time away from work due to disabilities, serious medical conditions, or pregnancy would naturally show lower activity levels in any system that measures output. By using those metrics as a primary filter, Meta's system would inevitably flag these workers as candidates for elimination, regardless of their actual job performance or value to the company.
This spring, Meta announced it would reduce its global workforce by roughly 10 percent—approximately 8,000 employees—as part of a broader restructuring. The layoffs began in May and continued in subsequent rounds. The 26 plaintiffs, who have chosen to remain anonymous, come from California, New York, Washington D.C., and several other states. They are arguing that the company violated both federal and state employment laws designed to protect workers from discrimination based on disability status or medical circumstances.
Meta has rejected the allegations outright. In a statement, the company said that workforce decisions were made by people, not machines. The company characterized the lawsuit as lacking merit and said it would defend itself vigorously. This framing—that humans, not algorithms, made the final calls—is a common response in cases like this, but it sidesteps the core claim: that the AI system shaped which employees were even presented as candidates for layoff in the first place.
The lawsuit arrives at a moment when artificial intelligence in human resources is drawing intense scrutiny from regulators and legal experts. Many large companies now use AI tools to help with hiring, performance reviews, promotions, and workforce planning. The appeal is obvious: automation promises speed, consistency, and the appearance of objectivity. But researchers and employment lawyers have warned repeatedly that poorly designed algorithms can encode discrimination if they rely on historical data or metrics that don't account for protected circumstances—like medical leave, disability accommodations, or pregnancy.
What makes this case potentially significant is its scope and its timing. If it proceeds to trial, it could become one of the most closely watched legal challenges involving AI in employment decisions. The outcome might reshape how technology companies and other large employers think about automating layoffs and other high-stakes workforce decisions. Courts could demand stricter compliance with anti-discrimination laws, or require companies to audit their algorithms before deploying them at scale. Alternatively, the case could be settled quietly, leaving the broader questions unresolved.
For now, Meta faces the task of explaining how its system worked, what data it used, and whether anyone at the company considered the possibility that measuring productivity alone might systematically harm workers whose circumstances made lower output inevitable. The 26 plaintiffs are betting that a jury will see what they see: not a neutral tool, but a mechanism that turned medical vulnerability into a reason to fire.
Notable Quotes
Meta said workforce management and organisational decisions were made by people rather than artificial intelligence— Meta statement
The Hearth Conversation Another angle on the story
Why does it matter whether humans or machines made the final decision if the algorithm shaped which people were even considered?
Because it changes what the company is responsible for. If a person chose to fire someone, that person can be questioned about their reasoning. But if an algorithm filtered the list first, the company can claim it was just following the numbers—even though someone had to decide which numbers to measure.
What would have been a fairer way to handle layoffs at that scale?
That's the hard question. You could exclude people on medical leave from the evaluation entirely. You could measure different things—client impact, team feedback, skills needed for the future. Or you could accept that layoffs will always hurt some people more than others and try to soften the blow through severance or retraining. But using a metric that automatically penalizes anyone who's been sick? That's hard to defend.
Does Meta have a point that people made the final decisions?
Technically, yes. But it's like saying a loaded die is fair because a human rolled it. The system was designed to flag certain workers. A person then approved those flags. The human didn't really choose—they just rubber-stamped what the machine suggested.
What happens if Meta wins this case?
Then companies get a green light to keep using productivity metrics in layoffs, even knowing they'll hurt disabled workers and people on medical leave. The message would be: as long as you can say a human signed off, the algorithm is fine.
And if the plaintiffs win?
Then Meta has to explain what it was measuring and why. Other companies will have to audit their systems before using them for layoffs. It might slow things down, but it could prevent a lot of harm.