AI fired me. Is that legal?

When 26 Meta employees were let go from their jobs this past April, they didn't go quietly into the night. They turned to legal counsel, filing a lawsuit against the company for wrongful termination. They alleged that Meta's AI algorithms unfairly "selected" them for termination in violation of a whole host of state and federal employment and disability laws. (Does 1 Through 26 v. Meta Platforms Inc. (N.D. Cal., No. 3:26-cv-07122, 7/13/26).)

The terminations were part of a downsizing effort that would eliminate approximately 8,000 positions at the company pursuant to a Reduction in Force (RIF), but Meta's algorithms, the plaintiffs contended, targeted employees who had asked for or taken protected leave. According to their complaint, "within the twenty-four months preceding the RIF" each of the named plaintiffs "took, requested, or was approved to take statutorily protected leave; attempted to take protected leave and suffered interference; or requested or received a reasonable accommodation for a disability."

This use, the plaintiffs alleged, violated the Family and Medical Leave Act and failed to meet the accommodation requirements of the Pregnant Workers Fairness Act and the Americans with Disabilities Act. The protected leaves, they argued, were intentionally used as a "negative factor" in the AI analysis. Although the district court denied the plaintiffs' motion for a temporary restraining order, finding that any potential harm could be remedied through the arbitration process (which the plaintiffs had separately initiated), the issues raised in their complaint merit serious consideration.

AI employment bias cases

Most AI bias cases have focused on employers' hiring and recruitment activities. They have looked at how the technology may unfairly filter out applicants who are members of classes protected under Title VII of the 1964 Civil Rights Act. In Mobley v. Workday, Inc. 740 F. Supp. 3d 796 ND CA 2024, the plaintiff asserted that Workday's algorithmic decision-making tools discriminated against him and job applicants who were African American, over age 40, and/or disabled.

The court denied Workday's argument that it was an "agent" and could therefore not be held directly liable. The complaint, it held, "plausibly alleges that Workday's customers delegated their traditional function of rejecting candidates or advancing them to the interview stage to Workday." Workday's tool, the court said, was not simply implementing the employers' hiring criteria, but was actually "participating in the decision-making process by recommending some candidates to move forward and rejecting others."

In the recent Meta case, the plaintiffs argued that when it used its AI tool to penalize them for taking leave, the employer violated their legal rights under FMLA and the ADA, neither of which require a showing of discrimination as a result of neutral acts. They also claimed that they were subjected to disparate treatment, retaliation and failure to accommodate under the law.

AI employment bias laws

In addition to their claims under federal law, the Meta employees charged the company with violating the California Fair Employment and Housing Act (FEHA). Specifically, they cited regulations adopted in 2025 that make it unlawful for an employer to use automated decision-making systems (ADS) or selection criteria that discriminate against either job applicants or employees based on protected categories under FEHA.

The workers argued that by factoring the AI system's output into its termination decisions, the algorithm was illegally stacking the deck against them. While they were on protected leave or had disability accommodations, they were unable to accumulate the output numbers required by the system.

A new law, enacted Sept. 4, 2026, should provide future California plaintiffs with even stronger support for claims such as those against Meta. SB 947, which takes effect July 1, 2027, further restricts employers' use of ADS to make disciplinary or termination decisions. Under new Labor Code section 1522, "An employer shall not rely solely on an ADS when making a disciplinary or termination decision."

If the employer primarily relies on ADS for such a decision, a human being must then corroborate the decision. They can do this by looking at the data that supported the ADS output or by considering other relevant information such as manager evaluations, peer reviews, employee work product or customer reviews. "If an employer cannot corroborate the ADS output, or the human reviewer has concluded that the ADS output is inaccurate, incomplete, or misleading, the employer shall not use the ADS output to make a disciplinary or termination decision."

Both laws define "automated decision system" broadly, encompassing more than just AI systems or large language models. The new section 1522 states that an ADS "may be derived from and or use artificial intelligence, machine learning, algorithms, statistics, or other data processing techniques." SB 947 defines it as "any computational process derived from machine learning, statistical modeling, data analytics, or artificial intelligence that issues simplified output, including a score, classification, or recommendation, that is used to assist or replace human discretionary decisionmaking and materially impacts natural persons."

Amendments to California's Consumer Privacy Act, which became effective Jan. 1, 2026, require employers using ADS to provide notice that such tools are used, offer opt-out unless human review can be done, and conduct risk assessments of automated hiring tools.

What it means in the workplace

These cases and laws establish a new level of accountability for AI use in employment decisions. Automated systems relied upon for hiring and termination decisions should be carefully vetted and monitored by employers; their use should be well documented and disclosed. A decision may be "data-driven," but the data can reflect past discrimination, missing context or faulty assumptions. A system that labels a worker a poor performer might overlook time off that was legally protected.

In order to defend against bias claims, employers should be able to demonstrate that their AI tools are job-related and address a specific business necessity. They should also be able to show that less discriminatory methods are unavailable.

Just because an AI tool is used for an employment action will not, by itself, establish liability. When Andon Market, a San Francisco retail store, used an AI tool this past August to fire an employee, the assumption was that the algorithm had operated independently. A further review showed that the AI tool had to be repeatedly prompted by human engineers before it was able to identify the employee handbook violation that justified the termination. Humans were an integral part of the entire process.

The red flags are clear. If a termination follows closely on the heels of protected activity such a disability leave, it is suspect. The same is true of a termination that overwhelmingly targets members of a protected class. If the employer cannot offer a plausible explanation for the termination, that too is problematic. And if that employer is unwilling to examine the basis for an AI tool's decision, all bets should be off.