US District Judge William Orrick in Oakland, California, indicated in a written order that he would not prevent Meta from executing layoffs starting July 22 while the validity of the workers’ unique legal claims is determined through private arbitration.
The judge noted that the employees failed to demonstrate that job loss constituted the “irreparable harm” necessary for an emergency order to stop the layoffs.
A Meta spokesperson chose not to comment. The company maintains it has done no wrong and asserts that layoffs were decided by human judgment.
In a joint statement, the plaintiffs’ attorneys mentioned that although Orrick rejected their appeal, he acknowledged that the lawsuit raises “serious questions” about Meta’s actions.
“The Court specifically noted it may revisit its decisions ‘based on any additional evidence the parties provide regarding whether and how AI was used’ in the workforce reduction,” they added.
In May, Meta informed nearly 8,000 employees, approximately 10% of its global workforce, that they would be losing their positions as the company intensifies its investment in AI.
The lawsuit, filed on Monday, alleges that Meta used AI tools to determine which jobs to cut based on productivity and AI token usage—thus disadvantaging those who were absent due to medical issues or caregiving. Additionally, the plaintiffs claimed that performance reviews, partly based on employees’ adoption of AI, played a role in the layoffs.
This case appears to be the first instance against a major US corporation challenging the alleged deployment of AI in layoffs.
The plaintiffs sought a temporary restraining order from Orrick to prevent Meta from finalizing layoffs while pursuing their claims in private arbitration.
Their request for a preliminary injunction, a more enduring temporary order, is currently pending. Orrick indicated on Friday that he might reconsider his position upon receiving more information regarding the layoffs.
During a Thursday hearing, the plaintiffs’ lawyers argued that in addition to losing their jobs and salaries, the employees risk losing valuable stock options and health insurance, jeopardizing their medical care for pregnancies and other conditions.
“There’s no second chance for bonding with a newborn or undergoing active medical treatment,” remarked attorney Barbara Cowan to Orrick.
Erin Connell, representing Meta, countered that the workers were only losing employer-subsidized insurance, not their overall coverage. Such damages are typically recoverable later if the plaintiffs succeed in arbitration, Connell stated.
The employees indicated that Meta’s agreements mandate individual arbitration for workplace disputes, but these do not apply to requests for temporary relief.
Most employees at large firms sign arbitration agreements, which generally require individuals to pursue workplace claims separately instead of collectively in court. Companies argue that arbitration is a quicker, less expensive alternative to litigation, while critics claim it often favors employers and discourages workers from filing claims.
While exceptions for temporary relief are common in arbitration agreements, they are typically utilized in cases involving alleged trade secret theft or solicitation of clients or employees, rather than layoffs of at-will workers.
The plaintiffs, who filed the lawsuit anonymously, include engineers, managers, researchers, and designers. They were informed about the layoffs in May, set to be finalized on July 22 for many and later in July or August for others, according to court documents.
Laid-off employees remain on the payroll but lost access to Meta systems as of May 20 and have not worked for the company since then, according to court filings.
They assert that Meta employed multiple internal AI-assisted systems to evaluate and rank employees on a termination list. These included a large language model assistant named “Metamate,” an employee-trained “second brain” that monitored communications and documents, along with a productivity score derived from analyzing keystrokes, screen content, emails, and browser history, as detailed in the lawsuit.
Meta did not suspend these systems during employees’ vacations and legally protected leave periods, causing the AI adoption scores used in layoff selections to decline, according to the plaintiffs.
