These workers say an AI hiring platform discriminated against them. But who should pay?

By Jenna Greene

Sept 21 (Reuters) - A lawsuit against Workday is testing the novel question of who bears responsibility when AI-powered hiring tools allegedly discriminate against Black and other applicants.

Last week, a group of job seekers asked a federal judge in San Francisco to certify their employment discrimination case against the human resources software provider as a class action.

The request marks a pivotal moment in the closely watched litigation, which stems from a Black plaintiff's claim that he was turned down from every job he applied for – more than 100 in total – with companies that used Workday's AI hiring tools.

While stories of AI agents breaking programming guardrails have dominated recent headlines, this suit stems from older technology that predates the current generative AI boom.

“We do not contend that the AI ‘went rogue,’” plaintiffs’ counsel Rod Cooks of Birmingham, Alabama-based Winston Cooks said via email. “Our concern is that the opposite may be true: an AI system can operate exactly as designed and trained while perpetuating patterns of past discrimination.”

Pleasanton, California-based Workday denies wrongdoing. A spokesperson said the company’s AI tools “only look at job qualifications—not protected traits like race, age, or disability,” adding that the plaintiffs “have not, and cannot, meet their burden to proceed as a class.”

The outcome could reverberate beyond Workday, which according to the company counts two-thirds of the Fortune 500 as customers.

As my Reuters colleague Daniel Wiessner has reported, numerous surveys have found that more than 80% of US ⁠employers are using AI in the hiring process. That includes using software made by Workday and other firms that can review large numbers of ​job applications and screen out applicants for a variety of reasons.

The case is “of critical importance,” employment litigator Gerald Maatman, who chairs the class action defense practice at Duane Morris and is not involved in the litigation, told me. If the plaintiffs succeed, he said, it “will create a blueprint for others to prosecute claims” against both AI software companies and their clients.

DISPARATE IMPACT

The litigation began in 2023, after co-lead plaintiff Derek Mobley was repeatedly rejected for financial, IT help-desk and customer service jobs, often within hours of applying.

Mobley’s resume includes that he graduated from Morehouse College, an all-male historically Black college and university, and shows his extensive employment history – alleged tip-offs to his race and that he is over age 40. According to the complaint, he also suffers from anxiety and depression.

Rather than suing the companies, however, his lawyers went after Workday.

The company says its Workday Recruiting software helps customers hire new workers – but that the product is highly customizable, and it’s up to each customer to decide how to sort, review and reject job applicants.

The plaintiffs contend the software itself contributes to discriminatory outcomes.

They point to statistical evidence to argue that the Workday tools disproportionately exclude African American, female, over-40 and disabled applicants from consideration, in violation of employment discrimination laws. They claim Workday uses data on a company's existing workforce to "train" the AI to screen for the best applicants without accounting for the existing bias that it may reflect.

In 2024, US District Judge Rita Lin in San Francisco allowed the case to move forward, though she rejected claims that Workday intended the tools to be discriminatory.

However, she found that the plaintiffs plausibly alleged the challenged technology resulted in a disparate impact.

The longstanding legal theory holds that employers can be liable for discrimination if otherwise neutral employment practices disproportionately affect a protected group and cannot be justified by business necessity.

Under the Trump administration, the federal government has stopped pursuing disparate impact cases, arguing that liability should instead turn on discriminatory intent. That doesn’t bar private litigants from asserting such claims, however – nor, Lin found, does it matter if a bot rather than a person is making an employment-related decision.

“Drawing an artificial distinction between software decision-makers and human decision-makers would potentially gut anti-discrimination laws in the modern era,” the judge wrote.

THE NEXT HURDLE

The key question now is whether the case can proceed as a class action.

The plaintiffs scored an earlier win on a similar but narrower question, when Lin in 2025 approved litigating the age bias claims as a collective action. The mechanism, most commonly used in employment cases involving wage-and-hour violations, is similar to a class action but requires people to proactively “opt in” to participate.

What the plaintiffs are asking for now is much broader. They want Lin to certify four subclasses -- for African Americans, women, people over 40 and people with disabilities who applied for jobs and were subjected to the challenged discriminatory screening process.

The scale is potentially enormous. Pointing to Workday’s own witness, the plaintiffs say that in 2024 alone, more than 356 million applications were submitted via Workday Recruiting, and more than 27 million job offers were extended.

The complaint seeks back pay, front pay, punitive damages and other relief.

Steven Nevolis, a labor and employment partner at Thompson Hine who is following the case but not involved, told me that the biggest hurdle for the plaintiffs to win class certification may be commonality. That is, can they show that the alleged discrimination stems from common features of Workday's technology rather than individualized hiring decisions made by its customers?

The would-be class members applied to different employers for different jobs, Nevolis noted, positioning defense lawyers to argue that “their individual experiences defeat commonality.”

The judge has scheduled a hearing on class certification for March 9, 2027.