Meta’s 8,000 layoffs: lawsuit alleges AI, not managers, picked employees for termination

Meta has rejected a serious accusation in a new lawsuit: that the company’s AI-driven layoff process for 8,000 employees selected workers based on health, disability, and leave status, not human manager judgment. The plaintiffs, 26 “Doe” employees, filed the complaint yesterday in US District Court for the Northern District of California, and they frame the core issue as process, not politics: Meta allegedly used internal artificial intelligence systems to construct a termination list.

According to the complaint, Meta did not “assemble the termination list through the considered judgment of managers who knew the work.” Instead, it allegedly relied on “a constellation of internal artificial-intelligence systems,” including a system referred to internally as “Metamate,” employee-trained “second-brain” agents, keystroke- and activity-monitoring data, AI-token-usage dashboards, and algorithmically assisted performance ranking and calibration. Put plainly, the allegation is that the layoffs were not just informed by AI, but operationalized through AI inputs and scoring mechanisms.

Meta denies that it used AI to terminate workers with disabilities and medical problems. That denial matters, because the plaintiffs are not just arguing that layoffs happened, they are arguing that the method used to choose who goes violated protected rights. The complaint also alleges that employees were graded on how much they used Meta’s AI tools. If those dashboards classified people by “stage of adoption,” as the lawsuit claims, then a metric intended to measure engagement could become, in effect, a proxy for whether someone was included on a layoff list.

The filing’s most specific operational claim is about internal categorization. The lawsuit says Meta’s internal dashboards classified employees by their stage of adoption of its artificial-intelligence tools, using categories such as “AI Native,” “AI First,” and “AI Enabled.” The plaintiffs’ theory is that these adoption-stage categories were among the inputs that helped determine inclusion on the termination list. This is a second-order problem for any large employer trying to roll out productivity or AI systems: once you start scoring behavior, you are one HR policy away from turning “tool usage” into “risk scoring.” Even if managers believe they are following a fair process, algorithmic inputs can quietly change what “performance” means.

This dispute lands in a regulatory and legal landscape that has been moving toward accountability for automated decision-making, especially when workers are at stake. In the United States, employment discrimination laws already prohibit adverse action tied to protected characteristics and protected leave. When AI is woven into performance ranking, monitoring, or staffing decisions, it can complicate the “who decided?” question. A human manager can say they relied on metrics, but plaintiffs can argue that the metrics encoded impermissible factors or magnified them through proxies. In the complaint, the plaintiffs are effectively making that argument by tying the layoff list to internal AI systems and monitoring data.

There is also a governance issue here. The plaintiffs’ complaint centers on the idea that manager judgment was displaced. That can become an internal board-level question fast: if leadership deployed AI tools for productivity, was there a parallel audit of how those tools might influence employment decisions? Were the models and scoring rubrics validated for disparate impact? Were the inputs, such as activity-monitoring and token-usage dashboards, tested for how they behave when employees have limitations, medical conditions, or different work patterns? The lawsuit does not ask “is AI smart,” it asks “is AI being used as a decision-maker,” and that distinction drives both legal exposure and reputational risk.

For executives, the immediate stake is litigation and the longer stake is what happens to enterprise decision-making norms. If plaintiffs persuade a court that layoffs were driven by algorithmic scoring systems, it can reshape how companies justify reductions in force, especially when AI is involved. It can also raise internal friction: managers may want to revert to pure human discretion, while policy teams may want to keep AI-assisted workflows but add guardrails, documentation, and oversight. For boards, this story is a reminder that deploying AI internally is not the end of the work. The “model lifecycle” now includes HR and employment, not just engineering and compliance.

Meta’s denial is part of the headline, but the complaint’s specificity is what changes the temperature. The allegation is concrete: a termination list allegedly generated through “Metamate,” “second-brain” agents, keystroke and activity monitoring, AI-token-usage dashboards, and algorithmically assisted performance ranking. If that allegation sticks, it would not just be a bad outcome for Meta; it would be a cautionary tale for every employer using internal AI systems to measure, rank, and optimize human work. The strategic question for peers is simple and urgent: when you scale AI inside the workplace, are you still in control of the decisions, or are you letting the tools write the operational playbook?