ABSTRACT
Law and legal compliance is the primary mechanism for risk management in any domain of industrial activity in the civilized world. In this chapter, we lay out the AI “control problem,” as well as other risks posed by AI with which the law must come to grips: AI is upending intellectual property regimes, enabling intrusive surveillance of private data, silently injuring vulnerable groups through algorithmic bias, posing challenges to democracy itself through weaponized bots interfering in the levers and pullies of government, and bringing about the end of truth with malicious deepfakes. The rule of law has never faced such a daunting challenge as the advent of AI. Let’s consider what laws might do to manage the risk of these AI-augmented powers. As law is a reflection of our appetite for risk, we contrast the precautionary principle with the more traditional cost-benefit analysis risk management approach. We recommend the former for dealing with the risks posed by AI. We take a deep dive into the problem of deepfakes to illustrate the challenge AI can pose to lawmakers. Issues of data privacy and security become increasingly urgent in light of the control problem because AI feeds upon, and could potentially weaponize, consumer data. The black-box nature of AI decision-making and issues posed by ingraining bias within AI systems also throw curve balls at traditional legal notions of liability and fault. The chapter concludes with a discussion of current approaches by advanced nation states to govern in the AI era.
