When AI Fires You: Algorithmic Discrimination and the Law
In Brief: AI systems are now making—or heavily influencing—hiring, performance, and termination decisions across U.S. workplaces. Workers on medical, parental, or disability leave are disproportionately harmed when automated scoring tools penalize legitimate absences. Federal and state laws still apply, and recent court rulings signal growing judicial scrutiny of algorithmic employment decisions.
In July 2026, 26 workers filed an anonymous lawsuit against Meta. Their allegation: that AI-powered productivity tools scored their performance while they were on approved medical, parental, or disability leave—then used those lowered scores to select them for termination in a mass layoff affecting nearly 8,000 employees.
This is widely reported as the first lawsuit against a major U.S. tech company to directly challenge AI-driven layoff decisions. It almost certainly will not be the last.
AI systems now influence every stage of the employment lifecycle—screening candidates before a human ever sees a résumé, monitoring keystrokes and browser activity, scoring performance, and flagging workers for layoffs. The technology moves fast. Legal accountability is catching up. And for workers in protected categories, the stakes could not be higher.
This post explains how algorithmic discrimination harms protected workers, what laws apply, how AI is reshaping pay equity, and what you can do if an automated system has affected your job.
How AI Has Taken Over Workplace Decision-Making
AI is no longer just a recruiting filter. Companies now deploy it across the entire employment relationship—from the moment a candidate submits an application to the moment a worker is selected for termination.
The systems at the center of the Meta lawsuit illustrate how far this has gone. According to the complaint, Meta used MetaMate, an internal large-language-model assistant, alongside an employee-trained “second brain” that tracked communications and documents, and a productivity-scoring tool that drew data from keystrokes, screen activity, browser history, messaging, and email. Meta has maintained that human managers made workforce decisions using neutral criteria—job level, historical performance ratings, and tenure—not AI, and not protected characteristics.
That defense highlights the central problem. When human decisions rely on data generated by automated systems, the line between algorithmic output and human judgment becomes difficult to locate—and nearly impossible for workers to challenge.
The scale of AI deployment is significant. According to a MyPerfectResume report, 73% of employers now use AI in hiring decisions, with half reporting that their tools automatically reject up to 50% of applications before any human review. More than 80% of U.S. employers, and virtually all Fortune 500 companies, currently use some form of AI screening in their hiring process.
Why Workers on Protected Leave Are Disproportionately at Risk
Many AI productivity and performance tools measure output continuously—without accounting for legitimate interruptions like medical leave, disability accommodations, pregnancy, or family caregiving. The result is a structural disadvantage built into the system itself, not necessarily into anyone’s intent.
In the Meta case, plaintiffs allege their AI-adoption metrics and productivity scores declined while they were lawfully away from work. Those lower scores, they contend, fed directly into the termination selection process. An algorithm that treats a worker on approved FMLA leave identically to one who is actively working will systematically penalize people with serious health conditions, disabilities, or pregnancy-related absences. The discrimination is encoded in the design.
This dynamic extends beyond layoffs. AI-driven systems increasingly personalize compensation—setting pay rates and bonus thresholds based on granular behavioral and performance data. When workers in protected categories generate less trackable “output” for legitimate reasons, their algorithmic scores—and corresponding pay—can fall accordingly. Two employees doing the same job at the same company may earn meaningfully different wages based entirely on AI-generated performance data, with no single discriminatory decision ever made.
The pattern is visible in industries far from Silicon Valley. Delivery and gig workers whose earnings are dictated by automated scorecards face situations where a minor metric decline—attributable to illness, injury, or a medical appointment—can eliminate bonus eligibility without any human review of the underlying cause.
Stanford University research reinforces the concern. A study following 3.4 million people submitting 4 million job applications across 1,700 positions found that an AI hiring tool can pass a standard bias audit at the aggregate level while still systematically screening out Black applicants and Asian applicants for specific roles. Applying the EEOC’s standard adverse impact threshold, the study found 26% of Black applicants and 15% of Asian applicants were affected in ways a standard audit would not have flagged.
What the Law Says—and Where Enforcement Falls Short
Existing federal and state laws apply to algorithmic employment decisions. The challenge is proving a violation when the algorithm itself is a proprietary trade secret.
Key legal protections include:
- Americans with Disabilities Act (ADA): Prohibits discrimination against qualified individuals with disabilities in hiring, firing, and compensation, and requires employers to provide reasonable accommodations.
- Family and Medical Leave Act (FMLA): Protects employees on approved leave from adverse employment actions, including termination.
- Title VII / Pregnancy Discrimination Act / Pregnant Workers Fairness Act: Prohibits discrimination based on sex, pregnancy, and related medical conditions.
- California FEHA: Provides broader state protections against disability and pregnancy discrimination, requiring employers to explore all reasonable accommodations before making an adverse employment decision.
Courts are beginning to take these claims seriously in the AI context. In Mobley v. Workday, a federal judge allowed discrimination claims to proceed under the ADA, California law, and federal anti-discrimination statutes. The plaintiff, Derek Mobley—a Black man over 40 with a disability—alleged he was rejected from more than 100 positions at companies using Workday’s AI screening platform, often within minutes of applying. Court filings show that approximately 1.1 billion applications were rejected using Workday’s tools during the relevant period. Workday denied wrongdoing, stating that its technology “looks only at job qualifications, not protected traits.”
In the Meta litigation, U.S. District Judge William Orrick declined to block the layoffs while acknowledging “serious questions going to the merits,” and separately required Meta to explain specifically why four visa-holding plaintiffs were selected—signaling that documentation of human decision-making carries real weight in court.
The EEOC has been direct on employer liability: employers are responsible for AI bias in tools supplied by third-party vendors. Delegating screening to an outside platform does not transfer legal exposure.
Colorado’s AI Act, effective June 2026, requires employers deploying high-risk AI systems to take reasonable care to protect consumers from algorithmic discrimination. California and New York City have enacted laws requiring bias testing of AI hiring tools. Compliance, however, remains inconsistent—and enforcement has not kept pace with deployment.
AI and Pay Discrimination: The Hidden Wage Gap
Algorithmic discrimination does not stop at who gets hired or fired. AI systems now personalize compensation in ways that can compound existing gender and racial wage gaps without generating any obvious paper trail.
Rather than applying a uniform wage, some AI compensation models calculate individualized pay offers based on behavioral indicators, location, work history, and projected acceptance rates. Workers who have taken protected leave, are managing disabilities, or have caregiving responsibilities may generate less trackable productivity data—and receive lower AI-generated compensation scores as a result.
What Workers Can Do Right Now
If you believe an AI system has influenced an adverse employment action against you, your response in the days and weeks that follow matters significantly.
Document everything. Keep records of approved leave, performance reviews, changes in productivity scores, and any communications about your role or compensation. Patterns matter in algorithmic discrimination claims—and courts have demonstrated they are willing to scrutinize them.
Know your rights. Employers are generally required to pause automated monitoring during approved leave or adjust scores to account for it. Failure to do so may constitute discrimination under the ADA, FMLA, Title VII, or applicable state law.
Demand transparency. In California and New York City, AI hiring tools are subject to bias-testing requirements that workers can invoke. You have a right to understand how decisions affecting your employment are being made.
Act quickly. Employment discrimination claims are subject to strict filing deadlines. Delaying consultation with an attorney can limit your legal options, sometimes significantly.
The Law Is Catching Up—But Workers Must Act
AI does not eliminate workplace discrimination. It can systematize it at scale, quietly and without obvious fingerprints. The Meta and Workday cases mark a turning point: courts are now scrutinizing AI-assisted employment decisions, and the legal frameworks workers need already exist.
What is required is workers who understand their rights, document their circumstances, and move before deadlines close.
If you were terminated, had your pay reduced, or experienced any adverse employment action that you believe was influenced by an AI system—and you were on medical leave, are pregnant, have a disability, or belong to any protected class—you may have a legal claim. Contact Helmer Friedman LLP for a confidential consultation.
Frequently Asked Questions
Can I sue my employer for using AI to fire me?
Yes, in certain circumstances. If an AI-assisted termination decision disadvantaged you because of a protected characteristic—such as disability, pregnancy, race, age, or sex—existing federal and state anti-discrimination laws may apply. Recent lawsuits against Meta and Workday have established that courts are willing to allow these claims to proceed. An employment attorney can help evaluate whether the specific facts of your situation support a viable legal claim.
Is AI-based pay discrimination illegal?
It can be. The Equal Pay Act, Title VII, and California’s Equal Pay Act prohibit pay disparities based on sex, race, and ethnicity. When AI compensation tools systematically assign lower pay to workers in protected categories—for example, those who have taken medical or parental leave—those disparities may constitute unlawful discrimination, even if no individual decision-maker intended to discriminate.
Are employers liable for AI discrimination if the tool came from a third-party vendor?
Yes. The EEOC has stated clearly that employers are responsible for vetting AI tools used in their hiring and employment processes, even when those tools are supplied by a third party. The Workday case also established a legal framework under which the vendor itself may face liability—but that does not eliminate the employer’s exposure.
What evidence do I need to challenge an AI-driven employment decision?
Documentation is critical. Records of approved leave, performance reviews, changes in scores or compensation, and any communications about your role or termination can help establish a pattern. Courts have shown they are willing to scrutinize algorithmic decision-making when workers can demonstrate a correlation between protected activity—such as taking FMLA leave—and adverse employment outcomes.
What is the deadline for filing an employment discrimination claim?
Filing deadlines vary depending on the type of claim and jurisdiction. Federal discrimination claims generally require filing a charge with the EEOC within 180 to 300 days of the discriminatory act. State law deadlines may differ. Acting promptly is essential—delaying consultation with an attorney can limit your options.