Following my own experience fighting pregnancy discrimination, I’ve come to think of employment discrimination as a black box. Workers who are denied promotions, passed over for opportunities, or selected for layoffs rarely know why. They see the outcome, but not the process. This makes it difficult to determine whether unlawful discrimination occurred and, if it did, to hold employers accountable under established federal and state antidiscrimination laws.
Artificial intelligence threatens to make that black box even darker.
After leaving Big Tech, with my once-held belief that we’re changing the world for the better lifted, I watched company after company announce waves of mass layoffs. My LinkedIn feed filled with stories of people hit by sudden and unexpected job loss, some in the midst of navigating incredibly vulnerable moments in life. One story I have never forgotten involved a Google employee who reportedly received notice of her layoff while she was in labor. At a time that should have been devoted to recovering from childbirth and caring for her newborn, she instead faced uncertainty about her family’s future, and, no doubt, the stress of job applications and interviews with a newborn in hand. I can empathize.
Those stories inspired me, more than two years ago, to write an article asking whether pregnant workers were being laid off by Big Tech companies at disproportionately higher rates than other employees. My point was never that every layoff involving a pregnant worker was unlawful. Rather, it was that we lacked the evidence necessary to know.
A growing problem
Today, that transparency problem has become even more urgent as artificial intelligence increasingly influences hiring, promotions, performance evaluations, and layoffs. Recent litigation against Meta showcases this. The lawsuit, filed in federal court by 24 former Meta employees, alleges that artificial intelligence played a substantial role in selecting them for layoffs, including employees on protected medical and family leave. The complaint alleges that Meta’s AI-driven layoffs drew on “inputs—performance ratings, calibration scores, productivity and output metrics, that, by design, cannot be accumulated by an employee who is on protected medical or family leave, or whose output is reduced by a disability.”
The path ahead is likely difficult for the plaintiffs, for many reasons. One is that a tech company like Meta will no doubt claim that its AI-driven employment practices are proprietary information.
Washington, my home state, has an opportunity to lead on this challenge as the home to many tech companies advancing AI development, startups and Big Tech alike. Earlier this month, an AI task force administered by the state attorney general published its final report recommending the development of worker-centered principles governing the use of AI in employment contexts, noting that while AI technologies “can improve efficiency, they also introduce risks of bias, inequity, and oversurveillance.” Although 2SHB 1833, a bill in the state House of Representatives calling for the establishment of a workplace AI advisory group, did not pass during the 2025 legislative session, the recommendation need not end there. Whether through legislative action or by Governor Bob Ferguson’s convening a multidisciplinary work group, Washington has the opportunity to bring workers, employers, technologists, and civil rights experts together to develop transparent, worker-centered principles for the use of AI in employment decisions.
The need for worker-centered AI principles became even clearer to me this summer as a participant in Seattle University School of Law’s Summer Initiative for Technology, Innovation, and Ethics. There, the investigative journalist Hilke Schellmann, author of The Algorithm: How AI Decides Who Gets Hired, Monitored, Promoted, and Fired and Why We Need to Fight Back, explained that many AI hiring and employment tools receive remarkably little testing before deployment. Furthermore, because builders of such tools often treat these systems as proprietary, employers who deploy them may not fully know how they were developed, what data they were trained on, or how rigorously they were evaluated for bias, if at all. Yet many organizations assume that if an AI product is commercially available, someone else has already evaluated it for fairness and bias. Too often, that assumption is misplaced.
This reflects a broader pattern within the technology industry. For years, Silicon Valley embraced the philosophy of “move fast and break things.” That mindset produced extraordinary innovation, but it also produced insufficiently vetted technologies that frequently outpaced regulators, courts, and even society’s understanding of their consequences. As Steve Tapia, a professor at the Seattle University law school, reminded our cohort at Summer Initiative, with AI the costs of “breaking things” are far greater because isolated errors can quickly become systemic harms affecting thousands of workers.
What transparency looks like
When employers rely on AI to make consequential employment decisions, meaningful transparency requires appropriate testing, disclosure, and human oversight. Companies should not be required to disclose trade secrets. But proprietary protections should not become a shield that prevents the public from obtaining the information necessary to evaluate whether AI-assisted employment decisions comply with antidiscrimination and other employment laws.
Meaningful transparency could include notifying employees when AI informs an employment decision, providing employees with a real opportunity to appeal AI-assisted decisions, and requiring employers to audit and report on whether large-scale workforce reductions driven by AI disproportionately affect protected groups.
Some argue that regulating artificial intelligence will slow innovation. Having spent more than a decade designing technology products, I believe the greater risk is allowing opaque AI systems to undermine it. Research consistently demonstrates that diverse and inclusive workplaces foster stronger innovation by bringing a wider range of perspectives to complex problems. When AI systems exclude talented workers because of hidden biases, the consequences extend beyond individual workers. Companies risk narrowing the diversity of perspectives that fuels innovation, weakening the quality of the products and services they create, and ultimately diminishing long-term economic growth. The goal is not to choose between innovation and regulation, but—as a recent Harvard Kennedy School paper observes—to pursue “a middle path that leverages technical innovation and smart regulation to maximize AI’s potential benefits while minimizing its risks.”
Transparency makes accountability possible. It enables employers to identify problems before they become systemic, allows regulators and courts to evaluate discrimination claims based on evidence rather than speculation, and gives workers an opportunity to understand—and, when appropriate, challenge—discriminatory employment decisions.
If we want AI to shape the future of work, we must open the black box.
