Twenty-six employees at Meta who were recently laid off made a bold claim in a lawsuit last week: They alleged the company had used AI to select people for layoffs, disproportionately impacting people who had taken leave or those with disabilities. The lawsuit identified a number of factors that the plaintiffs claimed had influenced the layoff selection process, from an internal AI assistant to dashboards that tallied employees’ AI token usage.
The judge in the case asked the company to explain its decision to terminate four of the plaintiffs whose immigration status was at risk, after denying the plaintiffs a temporary restraining order—which would have blocked Meta from completing the layoffs.
In a filing on Thursday, Linh Doan, a company HR leader explained its rationale for selecting those employees for termination—all of whom had visas sponsored by Meta—offering insight into how the company conducts layoffs. Doan, the director of HR Business Partner Enablement, repeatedly emphasized that these decisions were made by “human business leaders” and that AI was not involved in the selection process, disputing the claims put forth by the plaintiffs. She also denied that the leaders considered—or even had insight into—whether an employee was on leave, claiming layoff criteria “did not include leave status, leave history, disability status, accommodation requests, or any other protected characteristic.”
Doan revealed that the company first identified the parts of the organization that would be impacted by layoffs—down to a “cohort” of employees with a specific level and title—and then determined a business rationale for why the group should be included in the layoffs. Then leaders had to come up with “selection criteria” to evaluate the employees in that group.
“The selection criteria had to be objective and tied to the business rationale,” Doan said in the filing. “For example, if the business rationale was to retain top talent within a particular team, the selection criteria might identify employees with lower performance ratings or lower historical performance scores.” Beyond performance, the selection criteria could also include any of the following elements: job profile or level, tenure, location, specialized skills or experience, as well as what the company describes as spans and layers. (Spans refers to how many direct reports a manager has and layers captures where an employee sits in the organization relative to CEO Mark Zuckerberg.)
According to Doan, the selection criteria for a cohort was finalized before any individual employee was evaluated.“Decision-makers did not have the ability to deviate from those criteria,” she claimed.
The declaration also provided specific explanations for why each of the four employees with visas were laid off—and in three of those cases, the employees were selected for layoffs for performance-related reasons. For one of the plaintiffs in this group (described in the lawsuit as “Doe 4”), Doan claimed that the selection process for layoffs involved “two objective selection criteria applied in sequence,” which included job level and historical performance. The employee in question was on parental leave when he was laid off and had received a “Consistently Met Expectations” rating in his year-end 2025 review.
“Meta maintains a high-performance culture with seven possible performance ratings,” Doan said. “Consistently Met Expectations” is the lowest rating that does not fall below expectations. The decision to select employees meeting these criteria was part of a broader effort to reduce headcount at this level while retaining higher-performing employees and transitioning to a smaller, talent-dense organization.”
While Meta insisted that AI was not used to actually make layoff decisions, Doan’s declaration did not touch on whether an employee’s AI usage could influence their performance ratings, given those ratings were a key criterion for certain layoffs. Previous reporting has suggested that Meta does, in fact, consider employees’ AI adoption in its performance review process, and that the company has started using a tool called Checkpoint. (When reached by Fast Company, Meta was not immediately available for comment.) This lawsuit has also supported those claims, citing allegations from a director at the company who said “employee AI adoption was going to be a core assessment metric” as part of the Checkpoint program, the details of which were not clearly disclosed until after the May layoffs were announced. He claimed to monitor AI usage dashboard “almost daily” and that those dashboards did not account for employees being out on leave. This claim was echoed by several other plaintiffs, who said their AI usage scores had dropped significantly while out on leave.
As Fast Company has previously reported, many employees across corporate America report facing repercussions at work after they return from leave—or even while they are still on leave. Recent cases also suggest that taking parental leave can impact performance reviews: Joanne Barela, a former human capital consultant who worked at Deloitte, brought a case in which she claimed the company had penalized her in performance reviews after she went on parental leave. In 2024, Microsoft settled a case with the California Civil Rights Department for over $14 million, after the agency found that employees had allegedly faced retaliation and discrimination for taking leave, including lower bonuses and performance ratings.
As the lawsuit against Meta suggests, there are open questions about the unintended effects of tying performance metrics to an employee’s embrace of AI, even if companies are not explicitly outsourcing layoff decisions to AI. For employees who may already face bias—be it working parents or employees with disabilities—the consequences could be even greater.