Expert Warns AI Hiring Tools May Worsen Bias

The U.S. Equal Employment Opportunity Commission (EEOC) has clarified that employers can be held liable for discrimination caused by AI hiring tools, even if the tools are developed by a third-party vendor. This guidance falls under Title VII of the Civil Rights Act, which prohibits employment discrimination based on race, color, religion, sex, and national origin. The EEOC's stance treats AI-driven assessments like any other employee selection procedure. A landmark example of AI bias is Amazon's experimental recruiting tool, which was shut down after it was found to penalize resumes containing the word "women's" and downgrade graduates of two all-women's colleges. The system taught itself that male candidates were preferable because it was trained on a decade's worth of the company's predominantly male engineering resumes, illustrating how historical data imbalances can create biased outcomes. The core technical issue is often "Garbage In, Garbage Out." AI models learn from historical hiring data, and if that data reflects existing societal or company-level biases, the AI will not only replicate but can also amplify them at scale. This creates a feedback loop where the biased output is used to train future iterations of the model, further entrenching the initial bias. These systems can also lead to "proxy discrimination," where the AI uses seemingly neutral data points as stand-ins for protected characteristics. For example, an algorithm might use a candidate's zip code as a proxy for race or a gap in employment history as a proxy for gender, leading to discriminatory filtering without ever explicitly considering the protected attribute. In response, states are beginning to regulate these tools directly. New York City's Local Law 144, which took effect in 2023, mandates that employers using "automated employment decision tools" conduct annual independent bias audits and notify candidates about their use. Non-compliance can result in penalties of up to $1,500 per violation, per day. Other states are following suit. Illinois' Artificial Intelligence Video Interview Act requires employers to obtain consent from and provide explanations to candidates before using AI to analyze video interviews. Maryland law requires applicant consent for the use of facial recognition technology in interviews. For product leaders building in this space, mitigating these risks requires a multi-faceted approach. Strategies include ensuring training data is diverse and representative, conducting regular bias audits using frameworks like the EEOC's four-fifths rule, and designing "human-in-the-loop" systems where AI assists rather than replaces human judgment. The future of HR technology involves a shift toward more transparent and explainable AI. This includes AI agents that can orchestrate complex workflows, tools that structure and analyze unstructured data to identify skill gaps, and systems designed with fairness and accountability as core principles to avoid the legal and reputational damage of biased systems.