The machine that optimized the newsfeed might have also optimized who got to stay.
The Summary
- Twenty-six Meta employees filed a federal lawsuit alleging the company's internal AI systems disproportionately selected workers who took protected leave for termination during mass layoffs.
- The case centers on a legal question that will define the next decade of work: how do you prove an algorithm discriminated against you when the company won't show you the code?
- This isn't just about Meta. Every company using AI to make workforce decisions is watching to see if employees can pierce the black box.
The Signal
Meta built AI systems to organize the world's information. Now those systems allegedly helped decide which employees to cut. According to the lawsuit, the 26 plaintiffs believe Meta's internal AI tools flagged them for layoffs based on patterns like taking family leave, medical leave, or other protected time off. The company hasn't confirmed or denied the specific algorithmic claims, but the legal framework here matters more than Meta's defense.
Traditional employment discrimination cases have a playbook. You show a pattern of behavior, find the manager who made the call, pull the emails where they said the quiet part loud. AI-assisted layoffs break that model. There's no single decision-maker. There's a system that scores employees on dozens of variables, feeds those scores to managers, and the managers execute. Where does liability land?
"How do you prove an algorithm discriminated against you when the company won't show you the code?"
The plaintiffs face three linked problems:
- Proving the AI system existed and influenced termination decisions
- Demonstrating the system weighted protected characteristics in its scoring
- Connecting their specific layoffs to algorithmic output rather than human judgment
The lawsuit raises questions about algorithmic bias that courts haven't systematically answered yet. Can you compel discovery of training data? Do you have a right to see your algorithmic score? If the AI is a black box even to the company, are they still liable for its outputs? These aren't abstract questions anymore.
Meta's position is likely simple: managers made the final calls, AI was just one input among many, and the layoffs were performance-based. That's the standard defense. But if discovery shows that 90% of managers followed the AI recommendations without question, does the distinction between "AI-assisted" and "AI-driven" collapse?
The Implication
If the plaintiffs win, every company using AI for hiring, firing, or promotion decisions will need to audit their systems for disparate impact. That's expensive, but more importantly, it's technically hard. Most companies don't fully understand how their AI systems weight variables. They bought software from vendors who trained models on proprietary datasets. The chain of accountability dissolves.
Watch for two things. First, whether Meta tries to settle quickly and quietly to avoid setting precedent. Second, whether this case attracts regulatory attention from the EEOC or state labor boards. The legal infrastructure for algorithmic accountability is being built right now, one lawsuit at a time. The companies that prepare for transparency win. The ones that hope the black box stays closed are making a bad bet.