ABSTRACT
The capabilities of artificial intelligence (AI) systems have improved markedly over the past decade. This rapid progress has brought greater attention to longstanding concerns that advanced AI systems could cause catastrophic harm, up to and including human extinction. In principle, the prospect of tort liability could encourage AI developers to proceed with caution. But the current US tort liability system is not set up to handle the catastrophic risk posed by AI — since most of the expected harm from AI systems comes in truly catastrophic scenarios where compensation would not be feasible.
To address this, I propose a form of punitive damages designed to pull forward this expected liability into cases of practically compensable harm. To succeed in offering sufficient incentives for precaution, such punitive damages would need to be available even in the absence of human malice or recklessness. This paper also examines other legal mechanisms to ensure socially optimal investment in AI safety measures. Doctrinal changes considered include recognizing the training and deployment of advanced AI systems as an abnormally dangerous activity subject to strict liability, adopting a capacious concept of foreseeability for the purposes of evaluating proximate cause in cases of AI harm, and modifying the scope of legally compensable damages in cases involving loss of human life.
The paper also discusses further AI liability law changes that would require legislation, including requiring liability insurance for training and deployment of advanced AI systems, diverting a portion of punitive damages into an AI safety fund, and credibly announcing the availability of strict liability and punitive damages in advance of the first case in which they might be appropriate. The paper concludes by characterizing the limits of an ex post tort liability framework for limiting catastrophic AI risk and considers what complementary policies may be needed to plug the gaps left by such a framework.
Weil, Gabriel, Tort Law as a Tool for Mitigating Catastrophic Risk from Artificial Intelligence (January 13, 2024).
Leave a Reply