Harvard Scholar Says Universities Must Govern AI

Harvard Berkman Klein faculty associate and Emory law professor Ifeoma Ajunwa argues that legal debates over artificial intelligence must move beyond assigning blame after harm occurs toward earlier governance, design standards and public-interest regulation. In an edited interview with the Harvard Gazette, Ajunwa, who leads Emory’s AI and the Future of Work program and advises Fortune 500 companies on AI ethics, described three phases in how law has engaged with AI. The first centered on black-box systems that obscure responsibility, followed by questions about liability when algorithms mediate housing, hiring and credit. The current phase concerns broader efforts to govern AI from design through deployment. Ajunwa said the law must protect foundational principles such as equal opportunity, equal access and due process, especially as AI agents become more autonomous. She criticized America’s heavy reliance on post-event litigation and called for regulatory frameworks before harms escalate, including educational programs, industry norms, and limits on use cases that conflict with public values. Her work centers on AI in employment, notably her book The Quantified Worker, which argues that artificial intelligence and big-data systems have changed the traditional employment bargain. Workers exchange labor for wages but increasingly surrender personal data as well, often without compensation or meaningful consent. Drawing on social-contract thinking, she asks whether new technologies further disadvantage vulnerable workers or can be regulated to help them flourish. Ajunwa also discussed the concept of the sacrificial lamb, warning that AI’s early move fast and break things culture has disproportionately harmed already marginalized groups, particularly women and Black workers in hiring systems. She pointed to Mobley v Workday as a central case in AI employment discrimination litigation, noting efforts to certify it as a class action. She said society should not treat concentrated harm to specific groups as an acceptable cost of innovation, citing litigation over alleged harm to children as a reason to reject claims that equal sacrifice is permissible. She said AI governance has become difficult in a fragmented digital public sphere, where echo chambers and algorithmic amplification weaken consensus even when basic regulation is warranted. Ajunwa argued universities remain one of the few spaces for sustained moral and institutional deliberation, helping diverse communities debate technology, law and society when legislatures, markets and platforms fail to do so.