A groundbreaking lawsuit targeting Meta Platforms is exposing the serious challenges workers face when they believe artificial intelligence played a role in their termination — chief among them, actually proving it.
The case involves 26 former Meta employees who claim the tech giant used AI-powered tools to select them for layoffs, allegedly targeting people with disabilities or those who had taken medical or family leave. Legal experts say the lawsuit helps explain why, despite widespread predictions, a flood of employment discrimination cases involving AI has not yet materialized in U.S. courts.
In a ruling issued last week, U.S. District Judge William Orrick declined to block Meta from completing the terminations. He identified a central problem for workers making these kinds of claims, writing that the plaintiffs “were not in the rooms where it happened.” Without being present when decisions were made, employees often cannot gather the evidence needed to quickly win in court.
Adding to their difficulties, the former Meta employees — like the majority of American workers — signed arbitration agreements. Those agreements mean they cannot join together in a class-action lawsuit, cannot present their case to a jury, and cannot pursue a large public settlement. Instead, each person must resolve their dispute through a private arbitration process.
Companies tend to favor arbitration, describing it as a faster and less expensive option compared to traditional court proceedings. Worker advocates, however, argue the process tends to benefit employers and can discourage employees from pursuing claims at all. Because arbitration is confidential, damaging evidence uncovered in one case cannot easily be shared with other workers or the public.
Christine Webber, co-chair of the civil rights and employment practice at plaintiffs’ firm Cohen Milstein Sellers & Toll — which is not involved in the Meta case — explained the problem plainly: “Even if you establish that a particular system would produce discriminatory outcomes left and right, you have no way of sharing that information with other employees.”
Webber and other attorneys representing workers say these obstacles explain both the scarcity of high-profile AI employment lawsuits and why even this Meta case — which only seeks temporary relief — stands out as unusual.
One of the few comparable cases involves Workday, whose HR management software faces allegations of unlawfully filtering out job applicants based on race, age, and disability. Arbitration is not a barrier in that case because Workday does not have agreements with the job applicants of its client companies. Workday denies the allegations.
The arbitration agreements signed by the Meta workers do include a narrow exception allowing employees to seek a court order to stop an irreversible action. However, that exception is typically used in cases involving stolen trade secrets or the poaching of clients — not layoffs of at-will workers.
Judge Orrick denied the workers’ request for a temporary restraining order that would have halted the layoffs. He still must decide whether to issue a longer-lasting preliminary injunction that would restore the workers to their jobs while their individual arbitration cases proceed. He indicated he could reverse course and grant the injunction if the plaintiffs produce evidence “regarding whether and how AI was used in an improper manner.” A hearing is set for August 24, and either side may appeal his decision.
According to the lawsuit, Meta used AI tools that tracked employee productivity and how much workers used AI systems — a measurement known as AI token usage. Workers who were absent due to medical conditions or family caregiving responsibilities allegedly had lower scores as a result, making them targets for cuts.
The plaintiffs further allege that Meta relied on several internal AI systems, including a large language model assistant called “Metamate,” an employee-trained “second brain” that monitored workers’ communications and documents, and a productivity scoring system that analyzed keystrokes, screen content, emails, and browser activity.
Meta has pushed back firmly, stating in court filings and public statements that human beings made all decisions related to the nearly 8,000 layoffs announced earlier this year. The company denied using AI usage levels as a factor in identifying workers for termination or in conducting performance reviews. A Meta spokesman said Tuesday the company had no additional comment.
Judge Orrick noted that, because the plaintiffs could not offer evidence contradicting Meta’s account, he was obligated to take the company’s statements at face value.
Attorneys for the plaintiffs acknowledged last week that gathering evidence presents a steep challenge, and they called on current and former Meta employees with knowledge of how AI was used in the layoff process to come forward. “Meta holds virtually all the relevant information,” the legal team said in a joint statement.
