Regulating AI: Considerations that apply across domains

Abstract

Awareness that AI-based technologies have far outpaced the existing regulatory frameworks have raised challenging questions about how to set limits on the most dangerous developments (lethal autonomous weapons or surveillance bots, for instance). Under the assumption that the robotics industry cannot be relied on to regulate itself, calls for government intervention within the regulatory space-national and international-have multiplied. The various approaches to regulating AI fall into two main categories. A sectoral approach looks to identify the societal risks posed by individual technologies, so that preventive or mitigating strategies can be implemented, on the assumption that the rules applicable to AI, in say the financial industry, would be very different from those relevant to heath care providers. A cross-sectoral approach, by contrast, involves the formulation of rules (whether norms adopted by industrial consensus or laws set down by governmental authority) that, as the name implies, would have application to AI-based technologies in their generality. After surveying some domestic and international initiatives that typify the two approaches, the chapter concludes with a list of 15 recommendations to guide reflection on the promotion of societally beneficial AI.