A group of Meta workers sued the company, alleging that AI-supported layoff scoring produced discriminatory outcomes hidden behind nominal human review. Twenty-six employees say the process disadvantaged people on protected leave or with disabilities; Meta says humans made the final choices.
Twenty-six Meta workers joined the lawsuit. They allege that AI-assisted evaluation influenced layoff selection. The complaint says workers with disabilities or protected leave were disproportionately harmed.
Meta denies that AI made final termination decisions and says humans were responsible. The allegations had not been tested at trial. A human signature does not necessarily provide meaningful review if managers rely on opaque scores. Discrimination law applies to outcomes and processes regardless of whether software is involved.
Employment decisions reveal whether organizations can explain, audit and challenge automated recommendations when livelihoods and protected rights are at stake. The case may turn on governance more than the label 'AI.' An employer needs to show what the system measured, whether protected circumstances distorted the data, who could override it and whether reviewers had enough information and time to exercise independent judgment.
Audit evidence can include model inputs, override rates, validation results and communications about the tool's purpose. The filing presents plaintiffs' allegations, not findings of fact, and the exact model design and decision chain are not public. The evidence standard here is to separate what reporting establishes, what officials or parties claim and what remains unresolved.
Watch discovery requests for scoring records and human-override data. Watch whether the court allows collective claims and how Meta documents accommodation and leave decisions. Those developments will determine whether today's announcement changes durable conditions or remains an incomplete first account.