Behind the overnight layoffs of 8,000 people, Meta was exposed to using AI to "screen employees", 26 employees filed a collective lawsuit: AI Token has become an assessment item, and employees on vacation have become the "hardest hit area"

Nowadays, AI is increasingly involved in recruitment, performance evaluation, even code writing and customer service — but what happens when AI starts deciding who gets laid off?

Recently, Meta has been pushed into the spotlight over this exact issue:

This week, 26 Meta employees filed a class-action lawsuit in the U.S. District Court for the Northern District of California, accusing the company of relying heavily on AI systems to generate layoff lists during its massive workforce reduction this year. These algorithms were designed to favor employees with high work output and active AI usage, while overlooking those on medical leave, maternity leave, parental leave, and employees with disabilities — ultimately resulting in these marginalized workers being added to the layoff roster.

26 Employees File Collective Suit: AI Layoff Algorithm "Inadvertently Targets" Workers on Leave

The incident traces back to May this year. At that time, to further optimize its organizational structure and accelerate its AI transformation, Meta announced it would cut around 8,000 employees, accounting for roughly 10% of the company's total workforce.

Back then, Meta CEO Mark Zuckerberg even thanked departing employees for their contributions in an internal memo, stating that AI is "the most impactful technology of our generation," and that the company must adopt a more flexible, efficient organizational form to compete with a cohort of AI-native startups.

However, while the public was still discussing how Meta would go "all in on AI," a group of employees who received layoff notices began to suspect that it might not just be the management deciding their fate — AI could be the one calling the shots.

This week, 26 current and former Meta employees anonymously filed a lawsuit in the U.S. District Court for the Northern District of California. They stated that they received layoff notices as early as May this year, informed that their positions would be phased out starting July 22. As of the filing of the lawsuit, the 26 individuals were still officially Meta employees, but their departure procedures were about to begin — so they are requesting the court to intervene immediately to halt the layoffs from proceeding.

These 26 employees share one common trait: they all legally applied for medical leave, maternity leave, parental leave, family care leave, or requested reasonable accommodations due to physical disabilities — they all maintain that these experiences put them at an inherent disadvantage in the AI scoring system.

Deciding "Who Stays and Who Goes" Based on Productivity and AI Token Usage?

According to the lawsuit, Meta did not rely entirely on managers' subjective judgment when compiling the layoff list; instead, it utilized an evaluation system composed of multiple internal AI systems.

These systems comprehensively analyze a large volume of employee work data, including:

• Work productivity;
• Work output;
• AI Native score;
• AI Token usage amount;
• Rolling performance scores over the past 12 months;
• Calibrated scores;
• AI-aided performance rankings, and more.

Beyond that, Meta is also accused of using numerous digital monitoring tools.

For example, an internal AI tool called Metamate analyzes employees' internal communication records; other systems continuously collect data on employees' keyboard inputs, computer usage, email exchanges, and browser history to monitor work status and generate productivity metrics. Meta even uses some AI agents to analyze employees' documents and daily communications, attempting to "replicate" employees' work output as part of performance evaluations.

All this data ultimately determines each employee's composite score, which influences whether they are included in the layoff list — put simply: AI scores first, humans decide later.

But the 26 employees argue that the problem lies precisely with this scoring system itself.

Multiple Plaintiffs: Received Layoff Notices While on Maternity or Medical Leave

In reality, the core of this lawsuit is not whether Meta used AI for layoffs, but whether the AI evaluation system is biased: many of the metrics adopted by this AI algorithm are impossible for employees on legitimate leave to meet.

For instance, when an employee is on maternity or medical leave, they naturally do not write code, attend meetings, or frequently use the company's internal AI tools — so they cannot accumulate productivity data, AI token consumption, and other such metrics. As the lawsuit states: "These scores and metrics, by design, are unfair to employees who are on medical or family leave, or whose work output has decreased due to disability."

The key issue is that when calculating these scores, Meta made no special adjustments for employees on leave, nor did it pause the algorithmic scoring — instead, it directly incorporated these scores as the basis for layoffs. As a result: employees who took legal leave were systematically judged by the algorithm as "poor performers."

Several employees shared their specific experiences in the lawsuit:

One female employee, who had been approved for maternity leave by the company, received a layoff notice just two days before her due date.

Another employee was approved for medical leave by Meta's designated medical institution due to a "serious health condition and disability," but his direct supervisor warned him: if he took leave now, he would very likely be included in the upcoming layoff list. In the end, this employee not only failed to receive "reasonable accommodation," but was also placed on the layoff roster.

Additionally, according to statistics from the lawsuit, among all 26 laid-off employees: roughly half took leave due to pregnancy, childbirth, or caring for family members; 8 women took maternity or pregnancy-related leave; 4 men took parental leave; and one woman took leave to care for a family member, then later applied for bereavement leave after a relative passed away.

They argue that what truly led to their unemployment was not insufficient ability, but that AI equated "taking legal leave" with "declining performance," and accuse Meta's layoff practices of violating multiple U.S. federal laws including the Family and Medical Leave Act, the Americans with Disabilities Act, the Pregnancy Discrimination Act, and the Pregnant Workers Fairness Act.

In response to these allegations, Meta quickly issued a statement.

A company spokesperson said the claims are "unfounded and inconsistent with the facts." Meta emphasized that the company's organizational adjustments and personnel management decisions have always been made by managers, not by AI. However, as of now, Meta has not specifically responded to public questions about how its internal AI tools are involved in the layoff process, or what proportion of the decision-making weight the algorithms hold.

So what's your take on this incident?

This article is from the WeChat official account "CSDN", compiled by Zheng Liyuan, and published with authorization from 36Kr.