Phoenix 24
Algorithms may have converted protected leave into poor performance.
Menlo Park | July 2026
Meta is facing a federal lawsuit from 26 employees who allege that the technology company used artificial-intelligence systems and digital productivity measurements to help select workers for dismissal. The plaintiffs claim the process disproportionately affected employees who had taken legally protected medical, parental or family leave.
The lawsuit was filed in federal court in Oakland, California, after Meta notified approximately 8,000 employees that their positions would be eliminated as part of a restructuring announced in May. The reduction represented about 10 percent of the company’s global workforce.
The 26 plaintiffs remain employed but are scheduled to leave the company beginning July 22. They are seeking a temporary court order preventing their dismissals while their discrimination claims proceed through individual arbitration, as required by their employment agreements.
According to the complaint, Meta relied on a network of internal tools that collected and evaluated information about employee activity, output and use of artificial intelligence. The plaintiffs allege that these systems contributed to performance rankings and lists identifying workers whose positions would be eliminated.
The disputed tools allegedly included Metamate, Meta’s internal generative-AI assistant, employee-trained digital agents, dashboards measuring AI-token consumption and systems monitoring workplace activity. The complaint also refers to productivity information derived from keystrokes, screen content, communications, documents, emails and browsing activity.
These allegations have not been proven in court. Meta rejects the central claim and says artificial intelligence did not make the decisions.
A company spokesperson described the accusations as lacking factual merit and stated that workforce-management and organizational decisions were made by people, not AI. Meta has not publicly released the complete internal methodology used to select employees for the May restructuring.
The dispute therefore centers partly on what it means for a decision to be made by a human when managers may receive rankings, recommendations or performance indicators generated through automated systems. A supervisor can formally approve a dismissal while still depending heavily on algorithmically processed information.
The employees argue that Meta’s measurements failed to distinguish inactivity caused by poor performance from inactivity resulting from legally authorized absence. Someone on maternity, medical or caregiving leave would naturally produce fewer messages, code contributions, documents and AI-tool interactions during that period.
According to the complaint, the systems treated this reduced activity as weaker output without adequately correcting for the employee’s protected status. The plaintiffs contend that absence from the digital workplace became a negative performance signal, even when federal or state law authorized the leave.
Every plaintiff had allegedly taken protected leave, requested an accommodation or disclosed a disability. Approximately half had been absent for pregnancy, parental responsibilities, caregiving or bereavement.
The group reportedly includes eight women who took maternity or pregnancy-related leave, four men who used parental leave and one woman who took leave to care for a relative before later receiving bereavement leave. Other plaintiffs cited medical conditions, disabilities or workplace accommodations.
One employee alleged that a manager discouraged him from taking approved medical leave by warning that an absence could place him at greater risk during the expected workforce reduction. The complaint says Meta did not adequately adjust his assessment to account for his condition.
The workers accuse the company of violating several employment protections, including the Family and Medical Leave Act, the Americans with Disabilities Act, the Pregnancy Discrimination Act and the Pregnant Workers Fairness Act. They also invoke state laws protecting workers against discrimination and retaliation.
Another component of the lawsuit concerns algorithmic bias. The plaintiffs allege that Meta failed to perform legally required evaluations of automated employment systems under rules adopted in California and New York City.
The legal theory includes the concept of disparate impact. Under that principle, an apparently neutral policy may be discriminatory when it imposes a disproportionate burden on a protected group and cannot be justified as necessary for the job.
The employees argue that productivity systems based on continuously accumulated activity may affect women and caregivers more heavily because they are statistically more likely to use pregnancy, parental or family leave. A rule does not need to mention gender explicitly to produce an unequal result.
Meta can challenge that argument by demonstrating that managers conducted individualized reviews, excluded protected absences from their assessments and reached decisions using legitimate organizational criteria. The evidence may include internal instructions, performance records, management communications and technical documentation explaining how the tools operated.
The case could become one of the first major United States legal tests involving the alleged use of generative AI and automated monitoring during a mass dismissal. Previous regulatory debates have focused more frequently on algorithms used to advertise jobs, filter applications or screen candidates.
Termination decisions create a different level of risk. They can remove salaries, healthcare coverage, immigration sponsorship, unvested equity and access to benefits at moments when workers may already be managing pregnancy, disability or medical treatment.
The plaintiffs say completing their dismissals before arbitration would cause damage that later financial compensation might not fully repair. Some could lose employer-sponsored healthcare during active treatment or postpartum recovery. Foreign workers could face immigration consequences if their legal status depends on continued employment.
The dispute emerges as Meta reorganizes its workforce around an aggressive artificial-intelligence strategy. The company has increased investment in data centers, computing infrastructure, models and autonomous agents while reducing conventional positions and redirecting thousands of employees toward AI-related work.
Meta has projected capital expenditure of between $115 billion and $135 billion for 2026, substantially above its spending during the previous year. Management has described the layoffs as part of an effort to operate more efficiently and offset the cost of strategic investments.
The company is simultaneously encouraging employees to integrate AI into routine work. Internal systems are being developed to conduct research, produce software, navigate documents, prepare presentations and automate tasks previously completed by human teams.
That transformation creates a sensitive organizational contradiction. Employees may be expected to train and use the tools that measure their productivity, automate parts of their roles and potentially influence decisions about which positions remain necessary.
AI-token usage is particularly controversial as an employment metric. A low number may indicate that a worker has not adopted company technology, but it could also reflect the nature of the job, a protected absence or a deliberate decision not to use AI where human judgment is more appropriate.
High usage does not necessarily demonstrate high-quality performance. An employee could generate large quantities of automated output without producing reliable or valuable results. Measuring activity is substantially easier than determining whether the activity improved the organization.
Digital monitoring creates similar limitations. Keystrokes, screen movement and communication volume can show that someone is active, but they cannot independently measure strategic thinking, mentorship, creativity, complex problem-solving or the prevention of costly mistakes.
The legal question will not be whether Meta uses artificial intelligence internally. Companies are generally permitted to adopt tools that support productivity and workforce planning. The dispute concerns whether those systems produced discriminatory effects and whether human managers meaningfully corrected their limitations.
Meta’s defense that people made the final decisions may not end the inquiry. Courts and regulators increasingly examine the complete decision-making chain, including the data supplied to managers, the weight assigned to automated recommendations and the procedures available for correcting errors.
Human involvement can provide oversight, but it can also become a formal layer that approves a machine-generated conclusion without independently reviewing it. The quality of that review will be central to determining accountability.
The case also demonstrates why transparency matters in automated employment decisions. Workers should be able to understand which information affects their evaluation, how protected absences are excluded and whether they can challenge inaccurate data before an irreversible decision occurs.
AI systems can process organizational information at enormous scale, but scale does not guarantee fairness. A design error applied to one employee creates an individual injustice; the same error applied automatically across thousands of workers can become a structural employment problem.
Meta has denied using AI to determine the layoffs, and the plaintiffs must prove their allegations through the judicial and arbitration processes. Until internal records are examined, it remains uncertain whether the disputed technology selected employees, merely informed managers or played no decisive role.
The outcome could establish an important standard for other companies adopting automated management tools. Artificial intelligence may assist organizations in making employment decisions, but it does not remove their responsibility to respect disability accommodations, family leave and anti-discrimination law.
Cuando el algoritmo evalúa personas, la responsabilidad sigue siendo humana. / When the algorithm evaluates people, responsibility remains human.