The Meta AI Layoff Lawsuit Is a Wake-Up Call for Every TA Leader Using Algorithms

Twenty-six former Meta employees filed suit in California federal court this week claiming the company used AI-powered software that disproportionately targeted people with disabilities or who took medical leave when selecting workers for mass layoffs. According to HR Executive, the plaintiffs allege Meta assembled its termination list through a constellation of internal AI systems that did not consider the judgment of managers who knew the work. The plaintiffs, who worked across California, Florida, Illinois, New York, Pennsylvania and Washington state, all took, requested or were approved for protected leave within the past 24 months. They claim the AI penalized them for it.

Meta denies the claims. These claims lack merit and are not based on facts. Workforce management and organizational decisions were and are made by people, not AI, a Meta spokesperson said in a statement. But the lawsuit itself is the story. It is not a freak accident. It is the predictable outcome when companies deploy workforce algorithms without understanding what the system actually optimizes for.

What the Algorithm Learns When You Feed It Historical Performance Data

The lawsuit alleges that Meta relied on factors such as productivity and AI token usage when it cut thousands of jobs this year, disadvantaging people who missed work because of medical conditions or to care for family members. Here is the mechanism you need to understand. AI optimizes for patterns in historical data. If your high performers historically took less leave, the algorithm learns that leave-taking predicts low performance. It encodes discrimination as efficiency. The system is not malicious. It is doing exactly what it was trained to do: find correlations in past behavior and use them to predict future outcomes. The problem is that correlation is not causation, and protected leave is not a proxy for capability.

In our work with hiring teams, we see this pattern repeat. Organizations adopt algorithmic tools because they promise speed and scale, then discover too late that the tool has learned to replicate the biases baked into historical data. The difference between a hiring algorithm and a layoff algorithm is only timing. The liability is identical.

The Legal Standard Does Not Care Whether a Human or Algorithm Made the Decision

Meta’s defense is that people, not AI, made the decisions. That argument will not hold. The plaintiffs accuse the company of violating federal and state laws banning discrimination or retaliation against workers who have disabilities, take medical leave or are pregnant. They also claim Meta failed to test its AI systems for bias in violation of California and New York City laws. Under employment discrimination law, if your tool disproportionately harms a protected class, you must prove the criteria it uses are job-relevant and consistent with business necessity. Saying the AI recommended it is not a defense. You own the output.

This shifts AI hiring liability from theoretical to documented legal risk. Every TA leader using AI for any selection decision, hiring, promotion or layoff, now has legal precedent showing that algorithmic opacity does not shield you from accountability. If you cannot explain what your tool measures and why those measures are job-relevant and legally defensible, you are exposed.

Productivity Metrics Are Legitimate, Which Makes This Harder to Prove and Harder to Defend

The complicating factor is that productivity and output are legitimate business criteria. Meta is not accused of using facially discriminatory inputs like gender or disability status. The allegation is that the AI used facially neutral factors, productivity and token usage, that correlate with protected leave. This is harder to prove and harder to defend against. It requires you to show that the metric itself is a valid predictor of job performance and that less discriminatory alternatives do not exist.

Most TA teams cannot do that. They adopt vendor tools that promise to measure quality of hire or flight risk or cultural fit, and they trust the black box. They do not interrogate what the algorithm actually measures, whether it has been validated against job performance, or whether it has been tested for adverse impact. That ignorance is no longer defensible.

Your Vendor’s Validation Study Is Not Your Validation Study

If you are using AI to screen resumes, rank candidates, schedule interviews, predict retention or inform layoff decisions, can you articulate what the tool measures? Not what the vendor says it measures. What it actually measures, in your organization, with your data, for your roles. Can you show that the criteria correlate with job performance? Can you demonstrate that the tool does not produce disparate impact, or if it does, that no less discriminatory alternative exists?

If the answer is no, you are operating on faith. The Meta lawsuit is a reminder that faith is not a legal defense. The plaintiffs were notified in May that their jobs would be eliminated starting July 22. They are seeking a preliminary ruling blocking Meta from completing the layoffs while they pursue claims in private arbitration. Whether they win or lose, the precedent is set. Algorithmic workforce decisions are now subject to the same scrutiny as human ones, and the burden of proof is on you.

The lesson for TA is not to abandon AI. It is to stop deploying tools you cannot explain. Every algorithm makes trade-offs. It prioritizes some signals and ignores others. Those trade-offs encode values, whether you intended them or not. If you do not know what your tool optimizes for, you cannot know whether it is optimizing for the right things. And when the lawsuit arrives, you will not be able to defend it. The algorithm is not a shield. It is a mirror.

Original reporting: HR Executive.