Meta’s AI Tracked Their Keystrokes. Then It Put Them on the Layoff List.

Twenty-six former Meta employees filed suit in Oakland federal court on July 13, alleging that the company used AI-powered productivity monitoring tools to select them for layoffs while they were on legally protected leave. The case, brought under the FMLA, ADA, Pregnancy Discrimination Act, and Pregnant Workers Fairness Act, is the first major test of whether AI-assisted workforce selection constitutes discrimination when it intersects with protected status.

The plaintiffs, filing anonymously, describe a system in which Meta’s internal AI tools — productivity scores, keystroke logging, AI token usage dashboards, and the now-paused Model Capability Initiative (MCI) that captured mouse movements and screen content — created a persistent digital record of employee activity. When employees took protected leave for medical conditions, pregnancy, or disability accommodations, their monitored activity dropped. The lawsuit alleges that this drop was then used as a selection criterion for separation.

Meta denies the allegations. The company argues that the layoffs were part of a broader restructuring and that AI tools were used to assess team-wide productivity, not to target individuals on leave. An injunction to halt the separations was denied on July 17. The 26 employees are scheduled to be separated on July 22 — tomorrow.

The Legal Mechanism

The lawsuit does not claim that AI caused the layoffs in the abstract. It claims something more specific: that an AI system designed to measure productivity created a structural disadvantage for employees whose productivity metrics were suppressed by legally protected absence. The distinction matters. If the court finds that the AI tools functioned as a proxy for protected status — that is, that the system penalized employees for exercising legal rights — the ruling could establish a new liability framework for any company using AI in workforce decisions.

The legal landscape is already shifting around AI-assisted employment decisions. The EEOC has signaled that existing anti-discrimination frameworks apply to algorithmic selection tools, and the Mobley v. Workday case established that AI hiring platforms can face disparate impact claims. But no court has yet addressed the specific scenario in the Meta case: AI monitoring tools that generate the data used to select employees for separation while those employees are on protected leave.

The Monitoring-to-Selection Pipeline

What makes this case structurally different from previous AI employment lawsuits is the data pipeline. The plaintiffs allege that Meta’s AI tools did not merely assist human decision-makers — they generated the quantitative basis for the layoff decisions themselves. Keystroke frequency, screen activity, and AI tool usage were tracked, aggregated, and used to rank employees. When an employee went on medical leave, their activity metrics dropped to zero. When they returned, the lawsuit claims, the gap in their monitoring data was treated as a performance deficit rather than a legal right.

The MCI system, which Meta paused in June 2026 after internal concerns, was the most invasive component. According to the complaint, MCI captured mouse movements, keystrokes, and screen content at a granular level, creating a real-time productivity score that persisted even when employees were on approved leave. The system did not distinguish between an employee who was absent due to medical treatment and one who was simply idle.

What Separations Tomorrow Mean

The denial of the injunction on July 17 means the separations proceed as scheduled on July 22. The case will continue in discovery, but the immediate harm — the loss of employment — will have already occurred. For the 26 plaintiffs, the question is no longer whether they keep their jobs but whether the court finds that the selection process violated federal law.

The structural question extends far beyond Meta. Every company that uses AI to monitor employee productivity and then uses that data to make workforce decisions is now on notice: if the monitoring system does not account for legally protected absences, the selection process may be discriminatory by design. The case does not require proof that Meta intended to discriminate — only that the AI system produced discriminatory outcomes as a function of how it measured work.

The Challenger and Gallup data that this column has tracked shows a persistent gap between employer-reported AI-driven cuts (31%) and worker-identified AI layoffs (1%). The Meta case is where that attribution gap becomes a legal question. If a company’s AI monitoring system generates the data used to select employees for separation, and those employees were on protected leave, the system itself becomes the evidence. The outcome of this case will determine whether AI-assisted selection is treated as a neutral tool or as a structural mechanism that encodes discrimination into the workforce pipeline.