Twenty-six former Meta employees are suing the company over claims it used internal AI tools to rank and select workers for termination, according to The Verge AI, which cites the lawsuit and earlier reporting from Reuters. The core allegation: Meta’s scoring system never excluded people on parental or medical leave, so employees who took legally protected time off got disproportionately cut.
Meta denies it. “These claims lack merit and are not based on facts,” spokesperson Tracy Clayton told The Verge. “Workforce management and organizational decisions were and are made by people, not AI.”
📍 The Facts on the Ground
- Who: 26 former Meta employees, filing against Meta.
- When: The layoffs happened in May, part of Meta’s plan to cut 10 percent of staff, roughly 8,000 people.
- What they allege: Meta used a “constellation” of internal AI tools to score, rank, and select employees for the termination list.
- Which tools: Metamate (Meta’s internal AI assistant), employee-trained AI agents, and internal dashboards tracking AI token usage.
- The legal claim: Violations of federal and state laws that bar employers from firing workers for taking protected leave.
That token-usage detail is the one worth sitting with. If dashboards showing how much a person used internal AI tools fed into performance scoring, then someone on parental leave for three months shows up as a flat line. Not because they underperformed. Because they weren’t there. The lawsuit’s language is blunt on this: the scoring “in effect penalized the employees for exercising their legal rights to these leaves.”
🎯 Why This Matters
The status quo until now was fuzzy. Companies have quietly used analytics and performance data in layoff decisions for years. Nobody called it AI, and nobody had to explain the math. What’s changed is that the tooling got sophisticated enough to leave a trail, and that trail is now discoverable in litigation.
What stands out here is Meta’s defense. “Decisions were made by people, not AI” is going to be the standard response from every company facing this question. The trouble is that it’s a distinction without much weight if the humans were reading a ranked list an AI system produced. A person clicking approve on a sorted spreadsheet is still a person making a decision, technically. Courts are about to decide whether that’s good enough.
This is significant because it’s not really about Meta. It’s about whether “the model ranked them” survives contact with employment law.
⚙️ The Mechanics Nobody Planned For
Here’s the failure mode, explained plainly. AI scoring systems measure activity. Protected leave means the absence of activity. Unless someone explicitly writes a rule that says “exclude anyone who was out on leave during this scoring window,” the system treats a legally protected absence as a performance gap.
Nobody has to be malicious for that to happen. It’s a missing filter. And that’s exactly what makes it dangerous at scale, because a missing filter applies itself to 8,000 people at once with perfect consistency.
🧭 What to Watch
- Discovery. If this survives early motions, plaintiffs get to look inside how Metamate and those dashboards actually fed the process. That’s the part every HR and legal team in tech will be reading.
- Copycat suits. Meta isn’t the only company that ran mass layoffs with data-driven ranking in the last 18 months. A favorable ruling here opens the door.
- Audit requirements. Expect compliance teams to start demanding documented exclusion rules for protected classes before any AI-assisted workforce decision goes live.
If you build or operate internal scoring tools, the practical takeaway is small and boring: write down what the system excludes, and why. That document is either your defense or the plaintiff’s best exhibit.
The case is early. The precedent it sets won’t be. Full details are at the original source.